Scalable Video Coding Using Inter-Layer Prediction of Spatial Intra Prediction Parameters
The coding efficiency of scalable video coding is increased by substituting missing spatial intra prediction parameter candidates in a spatial neighborhood of a current block of the enhancement layer by use of intra prediction parameters of a co-located block of the base layer signal. By this measure, the coding efficiency for coding the spatial intra prediction parameters is increased due to the improved prediction quality of the set of intra prediction parameters of the enhancement layer, or, more precisely stated, the increased likelihood, that appropriate predictors for the intra prediction parameters for an intra predicted block of the enhancement layer are available thereby increasing the likelihood that the signaling of the intra prediction parameter of the respective enhancement layer block may be performed, on average, with less bits.
The present application is a continuation of U.S. patent application Ser. No. 16/180,292 filed Nov. 5, 2018, which is a continuation of U.S. patent application Ser. No. 14/666,658, filed Mar. 24, 2015, now U.S. Pat. No. 10,212,420, which is continuation of International Application PCT/EP2013/070492, filed Oct. 1, 2013, which claims priority from U.S. Provisional Application 61/708,201, filed Oct. 1, 2012, all of which are incorporated herein by reference in its entirety.
BACKGROUND OF THE INVENTIONThe present invention concerns scalable video coding.
In non-scalable coding, intra coding refers to coding techniques that do not reference data of already coding pictures, but exploit only data (e.g., reconstructed samples, coding mode, or symbol statistics) of already coded parts of the current picture. Intra-coded pictures (or intra pictures) are for example used in broadcast bitstreams in order to allow decoders to tune into a bitstream at so-called random access points. Intra pictures are also used to limit the error propagation in error-prone environments. In general, the first picture of a coded video sequence is necessitated to be coded as an intra picture, since here no picture are available that can be used as reference pictures. Often, intra pictures are also used at scene cuts where temporal prediction typically cannot provide a suitable prediction signal.
Furthermore, intra coding modes are also used for particular areas/blocks in so-called inter pictures, where they might perform better in terms of rate-distortion efficiency than inter coding modes. This is the often case in flat regions as well as in regions where temporal predictions performs rather poorly (occlusions, partially dissolves or fading objects).
In scalable coding, the concept of intra coding (coding of intra pictures and coding of intra blocks in inter pictures) can be extended to all pictures that belong to the same access unit or time instant. Therefore intra coding modes for a spatial or quality enhancement layer can also make use of inter-layer prediction from a lower layer picture at the same time instant to increase the coding efficiency. That means that not only already coded parts inside the current enhancement layer picture can be used for intra prediction, but also already coded lower layer pictures at the same time instant can be exploited. The latter concept is also referred to as inter-layer intra prediction.
In the state-of-the-art hybrid video coding standards (such as H.264/AVC or HEVC), the pictures of a video sequence are divided into blocks of samples. The block size can either be fixed or the coding approach can provide a hierarchical structure which allows blocks to be further subdivided into blocks with smaller block sizes. The reconstruction of a block is typically obtained by generating a prediction signal for the block and adding a transmitted residual signal. The residual signal is typically transmitted using transform coding, which means the quantization indices for transform coefficients (also referred to as transform coefficient levels) are transmitted using entropy coding techniques, and at the decoder side, these transmitted transform coefficient levels are scaled and inverse transformed to obtain the residual signal which is added to the prediction signal. The residual signal is generated either by intra prediction (using only already transmitted data for the current time instant) or by inter prediction (using already transmitted data for different time instants).
If inter prediction is used, the prediction block is derived by motion-compensated prediction using samples of already reconstructed frames. This can be done by unidirectional prediction (using one reference picture and one set of motion parameters), or the prediction signal can be generated by multi-hypothesis prediction. In the latter case, two or more prediction signals are superimposed, i.e., for each sample, a weighted average is constructed to form the final prediction signal. The multiple prediction signals (which are superimposed) can be generated by using different motion parameters for the different hypotheses (e.g., different reference pictures or motion vectors). For unidirectional prediction, it is also possible to multiply the samples of the motion-compensated prediction signal with a constant factor and add a constant offset in order to form the final prediction signal. Such a scaling and offset correction can also be used for all or selected hypothesis in multi-hypotheses prediction.
In current state-of-the-art video coding techniques, the intra prediction signal for a block is obtained by predicting samples from the spatial neighborhood (which was reconstructed before the current block according to the blocks processing order) of the current block. In the most recent standards various prediction methods are utilized that perform prediction in the spatial domain. There are fine-granular directional prediction modes where filtered or unfiltered samples of neighboring blocks are extended in a specific angle to generate the prediction signal. Furthermore, there are also plane-based and DC-based prediction modes that use neighboring block samples to generate flat prediction planes or DC prediction blocks.
In older video coding standards (e.g., H.263, MPEG-4) intra prediction was performed in the transform domain. In this case the transmitted coefficients were inverse quantized. And for a subset of the transform coefficients, the transform coefficient value was predicted using the corresponding reconstructed transform coefficient of a neighboring block. The inverse quantized transform coefficients were added to the predicted transform coefficient values, and the reconstructed transform coefficients were used as input to the inverse transform. The output of the inverse transform did form the final reconstructed signal for a block.
In scalable video coding also the base layer information can be utilized to support the prediction process for the enhancement layer. In the state-of-the-art video coding standard for scalable coding, the SVC extension of H.264/AVC, there is one additional mode for improving the coding efficiency of the intra prediction process in an enhancement layer. This mode is signaled at a macroblock level (a block of 16×16 luma samples). This mode is only supported if the co-located samples in the lower layer are coded using an intra prediction mode. If this mode is selected for a macroblock in a quality enhancement layer, the prediction signal is built by the co-located samples of the reconstructed lower layer signal before the deblocking filter operation. If the inter-layer intra prediction mode is selected in a spatial enhancement layer, the prediction signal is generated by upsampling the co-located reconstructed base layer signal (after the deblocking filter operation). For upsampling, FIR filters are used. In general, for the inter-layer intra prediction mode, an additional residual signal is transmitted by transform coding. The transmission of the residual signal can also be omitted (inferred to be equal to zero) if it is correspondingly signaled inside the bitstream. The final reconstruction signal is obtained by adding the reconstructed residual signal (obtained by scaling the transmitted transform coefficient levels and applying an inverse spatial transform) to the prediction signal.
However, it would be favorable to be able to achieve a higher coding efficiency in scalable video coding.
SUMMARYAccording to an embodiment, a scalable video decoder may be configured to reconstruct a base layer signal from a coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for at least a subset of blocks of the base layer signal for which the spatial intra prediction mode has been selected, reconstruct an enhancement layer signal from the coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for at least a subset of blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, having checking a neighboring block of the enhancement layer signal, neighboring a current block of the enhancement layer signal, as to whether same has been predicted using a spatial intra prediction mode; including, depending on the check, the intra prediction parameter of the neighboring block in a set of probably advantageous intra prediction parameters for the current block, or the intra prediction parameter of a block of the base layer signal, which is co-located to the current block, in the set of probably advantageous intra prediction parameters for the current block, and determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the set of probably advantageous intra prediction parameters.
According to another embodiment, a scalable video decoding method may have the steps of: reconstruct a base layer signal from a coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for at least a subset of blocks of the base layer signal for which the spatial intra prediction mode has been selected, reconstruct an enhancement layer signal from the coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for at least a subset of blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, having checking a neighboring block of the enhancement layer signal, neighboring a current block of the enhancement layer signal, as to whether same has been predicted using a spatial intra prediction mode; including, depending on the check, the intra prediction parameter of the neighboring block in a set of probably advantageous intra prediction parameters for the current block, or the intra prediction parameter of a block of the base layer signal, which is co-located to the current block, in the set of probably advantageous intra prediction parameters for the current block, and determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the set of probably advantageous intra prediction parameters.
According to another embodiment, a scalable video encoder may be configured to: encode a base layer signal into a coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for blocks of the base layer signal for which the spatial intra prediction mode has been selected, encode an enhancement layer signal into the coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, having checking a neighboring block of the enhancement layer signal, neighboring a current; block of the enhancement layer signal, as to whether same has been predicted using a spatial intra prediction mode; including, depending on the check, the intra prediction parameter of the neighboring block in a set of probably advantageous intra prediction parameters for the current block, or the intra prediction parameter of a block of the base layer signal, which is co-located to the current block, in the set of probably advantageous intra prediction parameters for the current block, and determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the set of probably advantageous intra prediction parameters.
According to still another embodiment, a scalable video encoding method may have the steps of: encode a base layer signal into a coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for blocks of the base layer signal for which the spatial intra prediction mode has been selected, encode an enhancement layer signal into the coded data stream by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, having checking a neighboring block of the enhancement layer signal, neighboring a current block of the enhancement layer signal, as to whether same has been predicted using a spatial intra prediction mode; including, depending on the check, the intra prediction parameter of the neighboring block in a set of probably advantageous intra prediction parameters for the current block, or the intra prediction parameter of a block of the base layer signal, which is co-located to the current block, in the set of probably advantageous intra prediction parameters for the current block, and determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the set of probably advantageous intra prediction parameters.
Another embodiment may have a computer program having a program code for performing, when running on a computer, the above decoding and encoding methods.
One aspect of the present application is that the coding efficiency of scalable video coding may be increased by substituting missing spatial intra prediction parameter candidates in a spatial neighborhood of a current block of the enhancement layer by use of intra prediction parameters of a co-located block of the base layer signal. By this measure, the coding efficiency for coding the spatial intra prediction parameters is increased due to the improved prediction quality of the set of intra prediction parameters of the enhancement layer, or, more precisely stated, the increased likelihood, that appropriate predictors for the intra prediction parameters for an intra predicted block of the enhancement layer are available thereby increasing the likelihood that the signaling of the intra prediction parameter of the respective enhancement layer block may be performed, on average, with less bits.
One aspect of the present application is that a better predictor for predictively coding the enhancement layer signal in scalable video coding may be achieved by forming the enhancement layer prediction signal out of an inter-layer prediction signal and an enhancement layer internal prediction signal in a manner differently weighted for different spatial frequency components, i.e. by forming a weighted average of the inter-layer prediction signal and the enhancement layer internal prediction signal at a portion currently to be reconstructed to obtain an enhancement layer prediction signal such that the weights at which the inter-layer prediction signal and the enhancement layer internal prediction signal contribute to the enhancement layer prediction signal vary over different spatial frequency components. By this measure, it is feasible to construe the enhancement layer prediction signal from the inter-layer prediction signal and the enhancement layer internal prediction signal in a manner optimized with respect to spectral characteristics of the individual contribution components, i.e. the inter-layer prediction signal on the one hand and the enhancement layer internal prediction signal on the other hand. For example, owing to the resolution or quality refinement based on which the inter-layer prediction signal is obtained from a reconstructed base layer signal, the inter-layer prediction signal may be more accurate at lower frequencies than compared to higher frequencies. As far as the enhancement layer internal prediction signal is concerned, its characteristic may be the other way around, i.e. its accuracy may be increased for higher frequencies compared to lower frequencies. In this example, the inter-layer prediction signal's contribution to the enhancement layer prediction signal should, by respective weighting, exceed the enhancement layer internal prediction signal's contribution to the enhancement layer prediction signal in the lower frequencies and decide the enhancement layer internal prediction signal's contribution to the enhancement layer prediction signal as far as the higher frequencies are concerned. By this measure, a more accurate enhancement layer prediction signal may be achieved, thereby increasing the coding efficiency and resulting in a higher compression rate.
By way of various embodiments, different possibilities are described to build the just outlined concept into any scalable video coding based concept. For example, the formation of the weighted average may be formed either in the spatial domain or the transform domain. Performance of the spectrally weighted average necessitates transformations to be performed on the individual contributions, i.e. inter-layer prediction signal and the enhancement layer internal prediction signal, but avoids for example spectrally filtering any of the inter-layer prediction signal and enhancement layer internal prediction signal in the spatial domain involving, for example, FIR or IIR filtering. However, performing the formation of the spectrally weighted average in the spatial domain avoids the detour of the individual contributions to the weighted average via the transform domain. The decision as to which domain is actually chosen for performing the formation of the spectrally weighted average may be dependent on whether the scalable video data stream contains, for the portion currently to be constructed in the enhancement layer signal, a residual signal in the form of transform coefficients or not: if not, the detour via the transform domain could be left off, while in case of an existing residual signal, the detour via the transform domain is even more advantageous since it allows for the transmitted residual signal in the transform domain to be added to the spectrally weighted average in the transform domain, directly.
One aspect of the present application is that information available from coding/decoding the base layer, i.e. base-layer hints, may be exploited to render the motion-compensated prediction of the enhancement layer more efficient by more efficiently coding the enhancement layer motion parameters. In particular, a set of motion parameter candidates gathered from neighboring already reconstructed blocks of the frame of the enhancement layer signal maybe enlarged by a set of one or more base layer motion parameters of a block of the base layer signal, co-located to the block of the frame of the enhancement layer signal, thereby improving the available quality of the motion parameter candidate set based on which the motion compensated prediction of the block of the enhancement layer signal may be performed by selecting one of the motion parameter candidates of the extended motion parameter candidate set and using the selected motion parameter candidate for the prediction. Additionally or alternatively, the motion parameter candidate list of an enhancement layer signal may be ordered dependent on base layer motion parameters involved in coding/decoding the base layer. By this measure, the probability distribution for selecting the enhancement layer motion parameter out of the ordered motion parameter candidate list is condensed so that, for example, an explicitly signaled index syntax element may be coded using less bits such as, for example, using entropy coding. Even further, additionally or alternatively, an index used in coding/decoding the base layer, may serve as a basis for determining the index into the motion parameter candidate list for the enhancement layer. By this measure, any signaling of the index for the enhancement layer may be avoided completely, or merely a deviation of the thus determined prediction for the index may be transmitted within the enhancement layer substream, thereby improving the coding efficiency.
One aspect of the present application is that scalable video coding may be rendered more efficient by deriving/selecting a subblock subdivision to be used for enhancement layer prediction, among a set of possible subblock subdivisions of an enhancement layer block by evaluating the spatial variation of the base layer coding parameters over the base layer signal. By this measure, less of the signalization overhead has to be spent on signaling this subblock subdivision within the enhancement layer data stream, if any. The subblock subdivision thus selected may be used in predictively coding/decoding the enhancement layer signal.
One aspect of the present application is that a subblock-based coding of transform coefficient blocks of the enhancement layer may be rendered more efficient if the subblock subdivision of the respective transform coefficient block is controlled on the basis of the base layer residual signal or the base layer signal. In particular, by exploiting the respective base layer hint, the subblocks may be made longer along a spatial frequency axis transverse to edge extensions observable from the base layer residual signal or the base layer signal. By this measure, it is feasible to adapt the subblocks' shape to an estimated distribution of the energy of the transform coefficients of the enhancement layer transform coefficient block in such a manner that, at an increased probability, each subblock will either be almost completely filled with significant, i.e. transform coefficients not having been quantized to zero, or with insignificant transform coefficients, i.e. only transform coefficients quantized to zero, while at a reduced probability any subblock has a similar number of significant transform coefficients on the one hand and insignificant transform coefficients on the other hand. Due to the fact, however, that subblocks having no significant transform coefficient may be signaled within the data stream efficiently, such as by use of merely one flag, and that subblocks almost completely filled with significant transform coefficients do not necessitate a waste of signalization amount for coding the insignificant transform coefficients which may be interspersed therein, the coding efficiency for coding the transform coefficient blocks of the enhancement layer is increased.
Embodiments are described further below in more detail with respect to the figures, among which:
Scalable video encoder 2 is configured to exploit redundancies between the versions 8a and 8b of video 4 reconstructible from base layer substream 6a without enhancement layer substream 6b on the one hand and both substreams 6a and 6b on the other hand, respectively. In order to do so, scalable video encoder 2 may use inter-layer prediction.
As shown in
Even if scalable video encoder 2 merely receives the original version of video 4, same may be configured to derive therefrom the two versions 4a and 4b internally, such as for example by obtaining the base layer version 4a by spatial down-scaling and/or tone mapping from higher bit depth to lower bit depth.
Before describing details of embodiments of the present application in more detail below, i.e. embodiments showing as to how the embodiments of
As shown in
In the same manner as
For the sake of simplifying the understanding of the following embodiments,
With respect to the following figures, it is described how a scalable video encoder or decoder such as those described above with respect to
In particular,
In accordance with the embodiments described now, the prediction of portion 28 uses both intra-layer prediction within the enhancement layer itself as well as inter-layer prediction from the base layer, so as to predict portion 28. However, the predictions are combined in a manner so that these predictions contribute to the final predictor of portion 28 in a spectrally varying manner so that, in particular, the ratio between both contributions varies spectrally.
In particular, the portion 28 is spatially or temporally predicted from an already reconstructed portion of the enhancement layer signal 400, namely any portion illustrated by hatching in
In order to obtain another prediction of current portion 28, inter-layer prediction is used. To this end, the base layer signal 200 is subject to a resolution or quality refinement at a portion 36 spatially and temporally corresponding to current portion 28 of the enhancement layer signal 400, so as to obtain a potentially resolution increased inter-layer prediction signal for current portion 28, with a refinement procedure being illustrated using arrow 38 in
Accordingly, two prediction contributions 34 and 39 exist for the current portion 28, and a weighted average of both contributions is formed to obtain an enhancement layer prediction signal 42 for the current portion 28 in a manner so that the weights at which the inter-layer prediction signal and the enhancement layer internal prediction signal contribute to the enhancement layer prediction signal 42 vary differently over spatial frequency components as illustrated schematically at 44 in
While the prediction signal 42 may be directly used by the enhancement layer signal 400 at current portion 28, alternatively a residual signal may be present within the enhancement layer substream 6b for current portion 28 resulting, by a combination 50 with the prediction signal 42 such as, for example, an addition as illustrated in
Summarizing the description of
Different possibilities exist for the weighting parameters 58 and how they control the formation of the spectrally weighted average in 41. For example, the weighting parameters 58 could merely signal one of two states for the current portion 28, namely one state activating the formation of the spectrally weighted average as described so far, and the other state deactivating the contribution of the inter-layer prediction signal 38 so that the final enhancement layer prediction signal 42 is, in that case, merely made up by the enhancement layer internal prediction signal 34. Alternatively, weighting parameter 58 for current portion 28 could switch between activating spectrally weighted average formation on the one hand and inter-layer prediction signal 39 forming the enhancement layer prediction signal 42 alone on the other hand. The weighting parameter 58 could also be designed to signal one of the three states/alternatives just mentioned. Alternatively, or additionally, weighting parameters 58 could control the spectrally weighted average formation 41 for the current portion 28 with respect to the spectral variation of the ratio between the weight at which prediction signals 34 and 39 contribute to the final prediction signal 42. Later on, it will be described that the spectrally weighted average formation 41 may involve filtering one or both of prediction signals 34 and 39 before adding same, such as using, for example, a high pass and/or low pass filter, and in that case weighting parameters 58 could signal a filter characteristic for the filter or filters to be used for the prediction of current portion 28. As an alternative, it is described hereinafter that the spectral weighting in step 41 could be achieved by a spectral component individual weighting in the transform domain, and accordingly in this case the weighting parameters 58 could signal/set these spectral component individual weighting values.
Additionally or alternatively, the weighting parameter for the current portion 28 could signal whether the spectral weighting in step 41 is to be performed in the transform domain or the spatial domain.
Again, it recalled that the prediction is performed by scalable video encoder and decoders, with using the prediction for reconstruction in decoder and encoder, respectively.
As described above, the parameters present within the enhancement layer substream 6b for the current portion 28 such as the residual information 59, or the weighting parameters 58, may signal as to whether the average formation 41 is performed within the transform domain as shown in
Later, it is described that, to obtain the layer-internal enhancement layer prediction signal, a difference signal may be computed, and managed, between an already reconstructed portion of an enhancement layer signal and the inter-layer prediction signal. Spatial prediction of the difference signal at a first portion collocated to the portion of the enhancement layer signal, currently to be reconstructed, from a second portion of the difference signal, spatially neighboring the first portion and belonging to the already reconstructed portion of the enhancement layer signal, may then be used to spatial predict the difference signal. Alternatively, temporal prediction of the difference signal at the first portion collocated to the portion of the enhancement layer signal, currently to be reconstructed, from a second portion of the difference signal, belonging to previously reconstructed frames of the enhancement layer signal may be used to obtain a temporally predicted difference signal. A combination of the inter-layer prediction signal and the predicted difference signal may then be used to obtain the layer-internal enhancement layer prediction signal, which is then combined with the inter-layer prediction signal.
With respect to the following figures, it is described how a scalable video encoder or decoder such as those described above with respect to
To explain this aspect, reference is made to
The situation shown in
Scalable video encoder and scalable video decoder may determine the set of predetermined neighboring blocks, here blocks 92, 94, out of the set of already coded neighboring blocks, here blocks 92 to 96, depending on, for example, a predetermined sample position 98 within current portion 28 such as the upper left sample thereof. For example, merely those already coded neighboring blocks of the current portion 28 may form the set of “predetermined neighboring blocks” which comprise sample positions immediately adjacent to predetermined sample position 98. In any case, the neighboring already coded/decoded blocks comprise samples 102 adjacent to current block 28 based on the sample values of which the area of current block 28 is to be spatially predicted. To this end, a spatial prediction parameter such as 56 is signaled in the enhancement layer substream 6b. For example, the spatial prediction parameter for current block 28 indicates a spatial direction along which sample values of samples 102 are to be copied into the area of current block 28.
In any case, at the time of spatially predicting current block 28, the scalable video decoder/encoder has already reconstructed (and encoded in case of the encoder) the base layer 200 using the base layer substream 6a, at least as far as the relevant spatially corresponding area of the temporally corresponding picture 22a is concerned such as, as described above, using block-wise prediction and using, for example, a block-wise selection between spatial and temporal prediction modes.
In
In order to allow for a reconstruction of the enhancement layer signal from the coded data stream with respect to block 28, for which here exemplarily the spatial intra-layer prediction 30 has been selected, an intra prediction parameter is used and coded within the bitstream as follows:
Intra prediction parameters are often coded using the concept of most probable intra prediction parameters, which is a rather small subset of all possible intra prediction parameters. The set of most probable intra prediction parameters may, for example, include one, two or three intra prediction parameters, while the set of all possible intra prediction parameters can for example include 35 intra prediction parameters. If the intra prediction parameter is contained in the set of most probable intra prediction parameters, it can be signaled inside the bitstream with a small number of bits. If the intra prediction parameter is not contained in the set of most probable intra prediction parameters, its signaling inside the bitstream necessitates more bits. Thus, the amount of bits to be spent for the syntax element for signaling the intra prediction parameter for the current intra-predicted block depends on the quality of the set of most probable, or probably advantageous, intra prediction parameters. By using this concept, on average a lower number of bits is necessitated for coding the intra prediction parameters, assuming that a set of most probable intra prediction parameters can be suitably derived.
Typically, the set of most probable intra prediction parameters is chosen in a way that it includes the intra prediction parameters of directly neighboring blocks and/or additional often used intra prediction parameters in form of, for example, default parameters. Since, for example, the main gradient direction of neighboring blocks is similar, it is generally advantageous to include the intra prediction parameters of neighboring blocks in the set of most probable intra prediction parameters.
However, if the neighboring blocks are not coded in a spatial intra prediction mode, those parameters are not available at the decoder side.
In scalable coding, it is however possible to use the intra prediction parameter of a co-located base layer block, and accordingly, in accordance with the aspect outlined below, this circumstance is exploited, by using the intra prediction parameters of a co-located base layer block in case of the neighboring blocks not being coded in an spatial intra prediction mode.
Thus, according to
First of all, a predetermined neighboring block such as block 92 or 94 of the current block 28 is checked whether same has been predicted using the spatial intra prediction mode, i.e. whether the spatial intra prediction mode has been selected for that neighboring block. Depending thereon, the intra prediction parameter of that neighboring block is included in the set of probably advantageous intra prediction parameters for the current block 28 or, as a substitute, the intra prediction parameter of the co-located block 108 of the base layer, if any. This process can be performed for each of the predetermined neighboring blocks 92 and 94.
If, for example, the respective predetermined neighboring block is not a spatial intra prediction block, then instead of using default predictors or the like, the intra prediction parameter of a block 108 of the base layer signal 200 is included in the set of probably advantageous inter prediction parameters for the current block 28, which is co-located to the current block 28. For example, the co-located block 108 is determined using a predetermined sample position 98 of current block 28, namely the block 108 covering the position 106 locally corresponding to the predetermined sample position 98 within the temporally aligned picture 22a of the base layer signal 200. Naturally, beforehand a further check could be performed as to whether this co-located block 108 within the base layer signal 200 is actually a spatially intra predicted block. In the case of
Thus, if block 108 co-located to current block 28 is spatially intra predicted, its intra prediction parameter signaled within the base layer substream 6a is used as a kind of substitute for any predetermined neighboring block 92 or 94 of current block 28, which does not have any intra prediction parameter due to same being coded using another prediction mode, such as a temporal prediction mode.
In accordance with another embodiment, in certain cases, the intra prediction parameter of a predetermined neighboring block is substituted by the intra prediction parameter of the co-located base layer block, even if the respective predetermined neighboring block is of the intra prediction mode. For example, a further check may be performed for any predetermined neighboring block of the intra prediction mode, as to whether the intra prediction parameter fulfills a certain criterion. If a certain criterion is not fulfilled by the intra prediction parameter of the neighboring block, but the same criterion is fulfilled by the intra prediction parameter of the co-located base layer block, then the substitution performed despite the very neighboring block being intra-coded. For example, the intra prediction parameter of the neighboring block could be replaced by the intra prediction parameter of the base layer block, if the intra prediction parameter of the neighboring block does not represent an angular intra prediction mode (but for example a DC or planar intra prediction mode), but the intra prediction parameter of the co-located base layer block represents an angular intra prediction mode.
The inter prediction parameter for the current block 28 is then determined based on a syntax element present in the coded data stream such as the enhancement layer substream 6b for the current block 28, and the set of probably advantageous intra prediction parameters. That is, the syntax element may be coded using less bits in case of the inter prediction parameter for the current block 28 being member of the set of probably advantageous intra prediction parameters than in case of being member of a remainder of the set of possible intra prediction parameters, disjoint to the set of probably advantageous intra prediction parameters.
The set of possible intra prediction parameters may encompass several angular directional modes, according to which the current block is filled by copying from the already coded/decoded adjacent samples by copying along the angular direction of the respective mode/parameter, one DC mode according to which the samples of the current block are set to a constant value determined based on the already coded/decoded adjacent samples such as, for example, by some averaging, and a plane mode according to which the samples of the current block are set to a value distribution following a linear function the slopes in x and y and the intercept of which are determined based on the already coded/decoded adjacent samples, for example.
The syntax element 114 signaled within the enhancement layer substream 6b for current block 28 could, for example, as illustrated in
Thus, the scalable video decoder may obtain or retrieve the syntax element 114 from the enhancement layer substream 6b, and the scalable video encoder may insert the syntax element 114 into the same, and the syntax element 114 is then used, for example, to index one spatial prediction parameter out of list 122. In forming list 122, the above described substitution may be performed according to which the predetermined neighboring blocks 92 and 94 are checked as to whether same are of the spatial prediction coding mode type. As described, if not, the co-located block 108 is, for example, checked as to whether same, in turn, is a spatially predicted block and if yes the spatial prediction parameter of the same, such as an angular direction 112, having been used for spatially predicting this co-located block 108, is included in list 122. If the base layer block 108 does also not comprise a suitable intra prediction parameter, list 122 may be left without any contribution from the respective predetermined neighboring block 92 or 94. In order to avoid list 122 being empty because, for example, both the predetermined neighboring blocks 92, 98 as well as the co-located block 108 lacking suitable intra prediction parameters because of, for example, being inter predicted, at least one of members 124 may unconditionally be determined using a default intra prediction parameter. Alternatively, list 122 may be allowed to be empty.
Naturally, the aspect explained with respect to
With respect to the following figures, it is described how a scalable video encoder or decoder such as those described above with respect to
Reference is made to
In particular, instead of signaling within the enhancement layer substream 6b subdivision information for current portion 28, the above description suggests selecting a subblock subdivision among a set of possible subblock subdivisions of current portion 28 such that the selected subblock subdivision is the coarsest among the set of possible subblock subdivisions which, when transferred onto the co-located portion 134 of the base layer signal, subdivides the base layer signal 200 such that within each subblock of the respective subblock subdivision the base layer coding parameters are sufficiently similar to each other. In order to ease the understanding, reference is made to
Accordingly, no subdivisioning information for block 28 needs to be signaled within the enhancement layer substream 6b, thereby increasing the coding efficiency. Moreover, the way of obtaining the subdivisioning as just outlined is applicable irrespective of any registration of the location of current portion 28 with respect to any grid or the sample array of base layer signal 200. In particular, the subdivision derivation also works in case of fractional spatial resolution ratios between base layer and enhancement layer.
Based on the subblock subdivisioning of portion 28 thus determined, portion 28 may be predictively reconstructed/coded. With regard to the above description, it should be noted that different possibilities exist in order to “measure” the coarseness of the different available subblock subdivisions of current block 28. For example, the measure of coarseness could be determined based on the number of subblocks: the more subblocks a respective subblock subdivisioning has, the lower its level. This definition has obviously not been applied in the case of
For the sake of completeness,
As described above, the selection just outlined may be implemented by traversing the possible subblock subdivisions according to some sequential order, such as an order of increasing or decreasing level of coarseness, and selecting that possible subblock subdivision from which possible subblock subdivision on the circumstance that within each subblock of the respective subblock subdivision the base layer coding parameters are sufficiently similar to each other, no longer applies (in case of using a traversal according to increasing coarseness) or firstly happens to apply (in case of using a traversal according to decreasing level of coarseness). Alternatively, all possible subdivisions may be tested.
Although in the above description of
Interestingly, however, the definition of“sufficiency” of similarity of base layer coding parameters within a certain subblock may be determined/defined merely over a subset of the base layer coding parameters. For example, the similarity may be determined based on the prediction modes only. Alternatively, prediction parameters further adjusting spatial and/or temporal prediction may form parameters which the similarity of base layer coding parameters within a certain subblock depends on.
Further, as already outlined above, to be sufficiently similar to each other, base layer coding parameters within a certain subblock may need to be equal to each other within the respective subblock completely. Alternatively, the measure of similarity used may need to be within a certain interval in order to fulfill the criterion of“similarity”.
As outlined above, the subblock subdivision selected is not the only quantity which may be predicted or transferred from the base layer signal. Rather, the base layer coding parameters themselves may be transferred onto the enhancement layer signal so as to derive, based thereon, enhancement layer coding parameters for the subblocks of the subblock subdivision obtained by transferring the selected subblock subdivision from the base layer signal to the enhancement layer signal. As far as motion parameters are concerned, for example, scaling may be used in order to take the transition from the base layer to the enhancement layer into account. Advantageously, merely those portions, or syntax elements, of the prediction parameters of the base layer are used to set the subblocks of the current portions subblock subdivision obtained from the base layer, which influence the similarity measure. By this measure, the fact that these syntax elements of the prediction parameters within each subblock of the subblock subdivision selected, are somehow similar to each other, guarantees that the syntax elements of the base layer prediction parameters, used to predict the corresponding prediction parameters of the subblocks of the current portion 308 are similar, or even equal to each other, so that, in the first case of allowing for some variance, some meaningful “mean” of the syntax elements of the base layer prediction parameters corresponding to the base layer signal's portion covered by the respective subblock may be used as the predictor for the corresponding subblock. However, it may also be that merely a part of the syntax elements contributing to the similarity measure, are used for predicting the prediction parameters of the subblocks of the subdivision of the enhancement layer in addition to the mere subdivision transferal itself such as merely predicting or pre-setting the modes of the subblocks of current portion 28 although the mode-specific base layer prediction parameters participate in the similarity measure determination.
One such possibility of using not only the subdivision inter-layer prediction from base layer to enhancement layer will now be described with respect to the next figures,
In accordance with the embodiment of
One such coding unit 308 of current picture 22b of the enhancement layer signal 400 is exemplarily assumed to be assigned to the inter-layer prediction mode and is illustrated using hatching. In a manner similar to
As already announced above, according to the embodiment of
In particular, according to the embodiment of
In particular, in accordance with the embodiment of
While in accordance with an alternative embodiment, the derivation of prediction details for performing the prediction within coding unit 308 could stop here, i.e. could be restricted to the derivation of the subdivision of coding unit 308 into prediction blocks and the assignment of these prediction blocks into ones coded using non-temporal or spatial prediction and to ones being coded using temporal prediction, it does not in accordance with the embodiment of
In accordance with the latter embodiment, all prediction blocks of coding unit 308 having the non-temporal prediction mode assigned thereto, are subject to non-temporal, such as spatial intra prediction while using prediction parameters derived from the prediction parameters of the locally coinciding intra-blocks of the base layer signal 200 as the enhancement layer prediction parameters of these non-temporal mode blocks. Such derivation may thus involve the spatial prediction parameters of the locally co-located intra-blocks of the base layer signal 200. Such spatial prediction parameters may, for example, be indications of an angular direction along which spatial prediction is to be performed. As outlined above, either the similarity definition by itself necessitates that the spatial base layer prediction parameters overlaid by each non-temporal prediction block of coding unit 308 are identical to each other, or, for each non-temporal prediction block of coding unit 308, some averaging over the spatial base layer prediction parameters overlaid by the respective non-temporal prediction block is used to derive the prediction parameters of the respective non-temporal prediction block.
Alternatively, all prediction blocks of coding unit 308 having the non-temporal prediction mode assigned thereto may be subject to inter-layer prediction in the following manner: first of all, the base layer signal is subject to a resolution or quality refinement to obtain an inter-layer prediction signal, at least within those areas spatially co-located to the non-temporal prediction mode prediction blocks of coding unit 308, and then these prediction blocks of coding unit 308 are predicted using the inter-layer prediction signal.
Scalable video decoder and encoder may either subject all of coding unit 308 to the spatial prediction or to the inter-layer prediction by default. Alternatively, scalable video encoder/decoder may support both alternatives, and signaling within the coded video data stream signals which version is used as far as the non-temporal prediction mode prediction blocks of coding unit 308 are concerned. In particular, the decision among both alternatives may be signaled within the data stream at any granularity such as, for example, individually for coding unit 308.
As far as the other prediction blocks of coding unit 308 are concerned, same may be subject to temporal inter prediction using the prediction parameters which may, just as it is the case for the non-temporal prediction mode prediction blocks, be derived from the prediction parameters of the locally coinciding inter-blocks. The derivation may thus, in turn, relate to motion vectors assigned to the corresponding portions of the base layer signal.
For all other coding units having any of the spatial intra prediction mode and the temporal inter prediction mode assigned thereto, same are subject to spatial prediction or temporal prediction in the following manner: in particular, same are further subdivided into prediction blocks having a prediction mode assigned thereto which is common for all of the prediction blocks within a coding unit and is, in particular, the same prediction mode as assigned to the respective coding unit. That is, differing from coding units such as coding unit 308 which has the inter-layer prediction mode associated therewith, coding units having the spatial intra prediction mode or having the temporal inter prediction mode associated therewith, are subdivided into prediction blocks of the same prediction mode, only, namely the prediction mode inherited from the respective coding unit from which they are derived by subdivision of the respective coding unit.
The subdivision of all coding units including 308 may be a quad-tree subdivision into prediction blocks.
A further difference between coding units of the inter-layer prediction mode, such as coding unit 308, and coding units of a spatial intra prediction mode or a temporal inter prediction mode is that when subjecting prediction blocks of spatial intra prediction mode coding units or temporal inter prediction mode coding units to spatial prediction and temporal prediction, respectively, the prediction parameters are set without any dependency on the base layer signal 200 such as, for example, by way of signaling within the enhancement layer substream 6b. Even the subdivision of coding units other than those having the inter-layer layer prediction mode associated therewith such as coding unit 308, may be signaled within the enhancement layer signal 6b. That is, inter-layer prediction mode coding units such as 308 have the advantage of a low bit rate signalization need: in accordance with an embodiment, the mode indicator for the coding unit 308 itself needs not to be signaled within the enhancement layer substream. Optionally, further parameters may be transmitted for coding unit 308 such as prediction parameter residuals for the individual prediction blocks. Additionally or alternatively, a prediction residual for the coding unit 308 may be transmitted/signaled within the enhancement layer substream 6b. While the scalable video decoder retrieves this information from the enhancement layer substream, a scalable video encoder in accordance with a current embodiment determines these parameters and inserts same into the enhancement layer substream 6b.
In other words, the prediction of the base layer signal 200 may be done using the base layer coding parameters in such a manner that same spatially vary over the base layer signal 200 in units of base layer blocks 304. The prediction modes available for the base layer may comprise, for example, spatial and temporal prediction. The base layer coding parameters may further comprise prediction mode individual prediction parameters such as an angular direction as far as spatially predicted blocks 304 are concerned, and motion vectors as far as temporally predicted blocks 304 are concerned. The latter prediction mode individual prediction parameters may vary over the base layer signal in units smaller than the base layer blocks 304, namely the aforementioned prediction blocks 306. To fulfill the above outlined requirement of sufficient similarity, it may be a requirement that the prediction mode of all base layer blocks 304 overlapping an area of a respective possible subblock subdivision, are equal to each other. Merely then may the respective subblock subdivision be shortlisted to get the selected subblock subdivision. The requirement may, however, be even more strict: It could be that the prediction mode individual prediction parameters of the prediction blocks, overlapping a common area of a respective subblock subdivision, have to be equal to each other, too. Merely subblock subdivisions which fulfill this requirement with respect to each subblock of this respective subblock subdivision and the corresponding area within the base layer signal may be shortlisted to get the finally selected subblock subdivision.
In particular, as briefly outlined above, there are different possibilities as to how to perform the selection among the set of possible subblock divisions. To outline this in even further detail, reference is made to
Obviously, it is favorable to try to avoid performing check 356 for all members of set 352, and accordingly, as shown in
With respect to the following figures, it is described how a scalable video encoder or decoder such as those described above with respect to
To explain the aspect, reference is made to
The situation shown in
Scalable video encoder and scalable video decoder may determine the set of predetermined neighboring blocks, here blocks 92, 94, out of the set of already coded neighboring blocks, here blocks 92 to 96, depending on, for example, a predetermined sample position 98 within current portion 28 such as the upper left sample thereof. For example, merely those already coded neighboring blocks of the current portion 28 may form the set of “predetermined neighboring blocks” which comprise sample positions immediately adjacent to predetermined sample position 98. Further possibilities are described with respect to
In any case, a portion 502 of a, according to the decoding/coding order, previously coded/decoded picture 22b of the enhancement layer signal 400, displaced from a co-located position of current block 28 by a motion vector 504 comprises reconstructed sample values based on which the sample values of portion 28 may be predicted such as by mere copying or by interpolation. To this end, the motion vector 504 is signaled in the enhancement layer substream 6b. For example, the temporal prediction parameter for current block 28 indicates a displacement vector 506 indicating the displacement of portion 502 from the co-located position of portion 28 in the reference picture 22b so as to be copied, optionally by interpolation, onto the samples of portion 28.
In any case, at the time of temporally predicting current block 28, the scalable video decoder/encoder has already reconstructed (and encoded in case of the encoder) the base layer 200 using the base layer substream 6a, at least as far as the relevant spatially corresponding area of the temporally corresponding picture 22a is concerned such as, as described above, using block-wise prediction and using, for example, a block-wise selection between spatial and temporal prediction modes.
In
In order to allow for a reconstruction of the enhancement layer signal from the coded data stream with respect to block 28, for which here exemplarily the temporal intra-layer prediction 32 has been selected, an inter prediction parameter such as a motion parameter is used and determined in any of the following manners:
A first possibility is explained with respect to
Further, a set 522 of one or more base layer motion parameters 524 of a block 108 of the base layer signal 200, collocated to portion 28, is gathered or generated from the base layer motion parameters. In other words, the motion parameter associated with co-located block 108 in the base layer is used to derive one or more base layer motion parameters 524.
The one or more base layer motion parameters 524, or a scaled version thereof, is then added 526 to the set 512 of motion parameter candidates 514 so as to obtain an extended motion parameter candidate set 528 of motion parameter candidates. This may be done in any of manifold ways such as simply appending the base layer motion parameters 524 at the end of the list of candidates 514, or in a different manner for which an example is outlined with respect to
At least one of the motion parameter candidates 532 of the extended motion parameter candidate set 528, is then selected and the temporal prediction 32, by motion compensated prediction, of the portion 28 is then performed using the selected one of the motion parameter candidates of the extended motion parameter candidate set. The selection 534 may be signalized within the data stream such as substream 6b for portion 28, by way of an index 536 into list/set 528, or may be performed otherwise as explained with respect to
As described above, it could be checked as to whether the base layer motion parameter 523 has been coded in the coded data stream such as base layer substream 6a using merging or not, and, if the base layer motion parameter 523 is coded in the coded data stream using merging, the adding 526 could be suppressed.
The motion parameters mentioned in in accordance with
The incorporation 526 of the motion parameters 523 used at a co-located portion 108 of the base layer signal into the set 528 of merging/motion vector candidates 532 enables a very effective indexing among the intra-layer candidates 514 and the one or more inter-layer candidate 524. The selection 534 may involve explicit signaling of an index into the extended set/list of motion parameter candidates in the enhancement layer signal 6b such as per prediction block, per coding unit or the like. Alternatively, the selection index 536 may be inferred from other information of the enhancement layer signal 6b or inter-layer information.
According to the possibility of
As to the motion parameters mentioned in
The ordering 544 may, alternatively, be done in accordance with a measure which measures the difference between the base layer motion parameter 523 of the base layer signal, relating to the block 108 of the base layer signal co-located to the current block 28 of the enhancement layer signal, and the base layer motion parameters 546 of spatially and/or temporally neighboring blocks 548 in the base layer. The determined ordering in the base layer is then transferred to the enhancement layer, so that the enhancement layer motion parameter candidates are ordered in a way that the ordering is the same as the determined ordering for the corresponding base layer candidates. In this regard, a base layer motion parameter 546 may be said to correspond to an enhancement layer motion parameter of a neighboring enhancement layer block 92, 94 when the associated base layer block 548 is spatially/temporally co-located to the neighboring enhancement layer block 92 and 94 that is associated with the considered enhancement layer motion parameters. Even alternatively, a base layer motion parameter 546 may be said to correspond to an enhancement layer motion parameter of a neighboring enhancement layer block 92, 94 when the neighborhood relation (left neighbor, top neighbor, A1, A2, B1, B2, BO or see
In order to explain this in more detail, reference is made to
In accordance with the alternative of
According to the possibility of
As to the motion parameters mentioned in
With respect to the following figures, it is described how a scalable video encoder or decoder such as those described above with respect to
The present aspect regards residual coding within the enhancement layer. In particular,
In accordance with the aspect described with respect to
That is, scalable video decoder/encoder would select a subblock subdivision among a set of possible subblock subdivisions on the basis of the base layer residual signal or the base layer signal. Then, a coding 414, or decoding, of the transform coefficient block 402 would be performed by applying the selected subblock subdivision. In particular, the positions of the transform coefficients 414 would be traversed in units of the subblocks 412 such that all positions within one subblock are traversed in an immediately consecutive manner with then proceeding to a next subblock in a subblock order defined among the subblocks. For a currently visited subblock, such as the subblock 412 for which the reference sign 412 has exemplarily been shown in 22
In order to nevertheless perform the evaluation 408, one or a combination the following options A to C may be used.
In particular, scalable video decoder/encoder may perform a transform 422 onto the base layer residual signal or the reconstructed base layer signal within portion 406 so as to obtain a transform coefficient block 424 of transform coefficients coinciding in size with a transform coefficient block 402 to be coded/decoded. An inspection of the distribution of the values of the transform coefficients within transform coefficient block 424, 426, may then be used in order to appropriately set the dimension of the subblocks 412 along the direction of horizontal frequencies, i.e. 428 and the dimension of the subblocks 412 along the direction of vertical frequencies, i.e. 432.
Additionally or alternatively, scalable video decoder/encoder may inspect all transform coefficient blocks of the base layer transform blocks 434 illustrated by different hatchings in
A further alternative which may additionally or alternatively be used in order to perform the evaluation 408 is to inspect the base layer residual signal or the reconstructed base layer signal within the spatial domain by use of edge detection 444 or determination of the main gradient direction, for example, deciding based on the extension direction of the edges detected or the gradient determined within the co-located portion 406 so as to appropriately set the subblock dimensions 428 and 432.
Although not specifically described above, it may be of advantage to, in traversing the positions of the transform coefficients and units of the subblocks 412, to traverse the subblocks 412 in an order which starts from the zero frequency corner of the transform coefficient block, i.e. the upper left corner in
In the following, different embodiments are described which could be combined with the aspects described above. The embodiments described below relate to many different aspects or measures to render scalable video coding even more efficient. Partially, the above aspects are described below in further detail, to present another derivate embodiment thereof, but with keeping the general concept. These below-presented descriptions may be used for obtaining alternatives or extensions of the above embodiments/aspects. Most of the embodiments described below, however, relate to subaspects which may, optionally, be combined with the aspects already described above, i.e. they may be implemented within one scalable video decoder/encoder concurrently with the above embodiments, but need not.
In order to be able to more easily understand the description brought forward now a more detailed embodiment for implementing a scalable video encoder/decoder suitable for incorporating the any of the embodiments and combination if embodiments, is presented next. The different aspects described below are enumerated by use of alphanumeric signs. The description of some of these aspects references elements in the figures described now, where, in accordance with one embodiment, these aspects could be commonly implemented. However, it should be noted that as far as the individual aspects are concerned, the presence of every of element in the implementation of a scalable video decoder/encoder is not necessary as far as every aspect is concerned. Depending on the aspect in question, some of the elements and some of the inter-connections may be left out in the figures described next. Merely the elements cited with respect to the respective aspects should be present in order to perform the task or function mentioned in the description of the respective aspect, wherein, however, sometimes alternatives do also exist especially when several elements are cited with respect to one function.
However, in order to provide an overview of the functionality of a scalable video decoder/encoder where the aspects subsequently explained may be implemented, the elements shown in the figure below are briefly described now.
The scalable video decoder comprises two block-based predictive decoders 80, 60 running in parallel and receiving substream 6a and 6b, respectively. As shown in the figure, a demultiplexer 40 may separately provide decoding stages 80 and 60 with their corresponding substreams 6a and 6b.
The internal construction of the block-based predictive coding stages 80 and 60 may, as shown in the figure, be similar. From an input of the respective decoding stage 80, 60 on, an entropy decoding module 100; 320, an inverse transformer 560; 580, an adder 180; 340 and optional filters 120; 300 and 140; 280 are serially connected in the order of their mentioning so that at the end of this serial connection a reconstructed base layer signal 600 and a reconstructed enhancement layer signal 360, respectively, are derivable. While the output of adders 180, 340 and filters 120, 140, 300 and 280 provide different versions of reconstructions of the base layer and enhancement layer signals, respectively, a respective prediction provider 160; 260 is provided in order to receive a subset or all of these versions and to provide, based thereon, a prediction signal to the remaining input of adder 180; 340, respectively. The entropy decoding stages 100; 320 decode from the respective input signal 6a and 6b, respectively, transform coefficient blocks entering inverse transformer 560; 580 and coding parameters including prediction parameters for prediction provider 160; 260, respectively.
Thus, prediction providers 160 and 260 predict blocks of frames of the video at the respective resolution/quality level and, to this end, same may choose among certain prediction modes such as a spatial intra-prediction mode and a temporal inter-prediction mode, both of which are intra-layer prediction modes, i.e. prediction modes solely depending on the data in the substream entering the respective level.
However, in order to exploit the aforementioned inter-layer redundancies, enhancement layer decoding stage 60 additionally comprises a coding parameter inter-layer predictor 240, a resolution/quality refiner 220 and/or a prediction provider 260 which, compared to prediction provider 160, additionally/alternatively supports inter-layer prediction modes which are able to provide the enhancement layer prediction signal 420 based on data derived from internal states of base layer decoding stage 80. The resolution/quality refiner 220 is to subject any of the reconstructed base layer signals 200a, 200b and 200c or the base layer residual signal 480 to a resolution or quality refinement to obtain an inter-layer prediction signal 380, and the coding parameter inter-layer predictor 240 is to somehow predict coding parameters such as prediction parameters and motion parameters, respectively. The prediction provider 260, may, for example, additionally support inter-layer prediction modes according to which reconstructed portions of the base layer signal, such as 200a, 200b and 200c, or reconstructed portions of the base layer residual signal 640, potentially refined to the increased resolution/quality level, are used as a reference/basis.
As described above, the decoding stages 60 and 80 may operate in a block-based manner. That is, frames of the video may be subdivided into portions such as blocks. Different granularity levels may be used to assign the prediction modes as performed by prediction providers 160 and 260, the local transformations by inverse transformers 560 and 580, filter coefficient selections by filters 120 and 140 and prediction parameter settings for the prediction modes by prediction providers 160 and 260. That is, the sub-partitioning of the frames into the prediction blocks may, in turn, be a continuation of the sub-partitioning of the frames into the blocks for which the prediction modes are selected, called coding units or prediction units, for example. The sub-partitioning of the frames into blocks for transform coding, which may be called transform units, may be different from the partitioning into prediction units. Some of the inter-layer prediction modes used by prediction provider 260 are described below with respect to the aspects. The same applies with respect to some intra-layer prediction modes, i.e. prediction modes which internally derive the respective prediction signal input into adders 180 and 340, respectively, i.e. solely base on states involved in the current level's coding stage 60 and 80, respectively.
Some further details of the blocks shown in the figure will become apparent from the description of the individual aspects below. It is noted that, as far as such descriptions do not specifically relate to the aspect with respect to which they have been provided, these descriptions shall be equally transferable to the other aspect and the description of the figure in general.
In particular, the embodiment for a scalable video decoder of
The techniques for intra coding of an enhancement layer signal used in the embodiment described next include multiple methods for generating an intra prediction signal (using base layer data) for enhancement layer blocks. These methods are provided in addition to methods that generate the intra prediction signal based on reconstructed enhancement layer samples only.
Intra prediction is part of the reconstruction process of intra coded blocks. The final reconstruction block is obtained by adding the transform-coded residual signal (which may be zero) to the intra prediction signal. The residual signal is generated by inverse quantization (scaling) of the transform coefficient levels transmitted in the bitstream followed by an inverse transform.
The following description applies to scalable coding with quality enhancement layers (the enhancement layer has the same resolution as the base layer, but represents the input video with a higher quality or fidelity) and scalable coding with spatial enhancement layers (the enhancement layer has a higher resolution, i.e., a larger number of samples, than the base layer). For quality enhancement layers, no upsampling of base layer signals is necessitated, such as in block 220, but a filtering, such as 500, of the reconstructed base layer samples may be applied. In case of spatial enhancement layers, an upsampling of the base layer signals is generally necessitated, such as in block 220.
The aspect described next supports different methods for using the reconstructed base layer samples (cp. 200) or base layer residual samples (cp. 640) for intra prediction of an enhancement layer block. It is possible to support one or more of the methods described below in addition to an intra-layer intra coding (in which only reconstructed enhancement layer samples (cp. 400) are used for intra prediction). The usage of a particular method can be signaled at the level of the largest supported block size (such as a macroblock in H.264/AVC or a coding tree block/largest coding unit in HEVC), or it can be signaled at all supported block sizes, or it can be signaled for a subset of the supported block sizes.
For all methods described in the following, the prediction signal can be directly used as reconstruction signal for the block, i.e. no residual is transmitted. Or the selected method for inter-layer intra prediction can be combined with residual coding. In a particular embodiment the residual signal is transmitted via transform coding, i.e., quantized transform coefficients (transform coefficient levels) are transmitted using an entropy coding technique (e.g., variable length coding or arithmetic coding (cp. 19b)) and the residual is obtained by inverse quantizing (scaling) the transmitted transform coefficient levels and applying an inverse transform (cp. 580). In a particular version, the complete residual block that corresponds to the block for which the inter-layer intra prediction signal is generated is transformed (cp. 726) using a single transform (i.e., the entire block is transformed using a single transform of the same size as the prediction block). In another embodiment, the prediction block can be further subdivided (e.g., using a hierarchical decomposition) into smaller block and for each of the smaller block (which can also have different block sizes) a separate transform is applied. In a further embodiment, a coding unit can be divided into smaller prediction blocks and for zero or more of the prediction blocks the prediction signal is generated using one of the methods for inter-layer intra prediction. And then, the residual of the entire coding unit is transformed using a single transform (cp. 726), or the coding unit is subdivided into different transform units, where the subdivision for forming the transform units (blocks to which a single transform is applied) is different from the subdivision for decomposing a coding unit into prediction blocks.
In a particular embodiment, the (upsampled/filtered) reconstructed base layer signal (cp. 380) is directly used as prediction signal. Multiple methods for employing the base layer for intra predicting the enhancement layer include the following method: The (upsampled/filtered) reconstructed base layer signal (cp. 380) is directly used as enhancement layer prediction signal. This method is similar to the known H.264/SVC inter-layer intra prediction mode. In this method a prediction block for the enhancement layer is formed by the co-located samples of the base layer reconstruction signal which might have been upsampled (cp. 220) to match the corresponding sample positions of the enhancement layer and might have been optionally filtered before or after the upsampling. In contrast to the SVC inter-layer intra prediction mode, this mode may be not only supported at a macroblock level (or largest supported block size), but at arbitrary block sizes. That means, the mode cannot only be signaled for the largest supported block size, but a block of the largest supported block size (macroblock in MPEG-4, H.264 and coding tree block/largest coding unit in HEVC) can be hierarchically subdivided into smaller blocks/coding units and the usage of the inter-layer intra prediction mode can be signaled at any supported block size (for the corresponding block). In a particular embodiment, this mode is only supported for selected block sizes. Then, the syntax element that signals the usage of this mode can be transmitted only for the corresponding block sizes, or the values of a syntax element that signals (among other coding parameters) the usage of this mode can be correspondingly restricted for other block sizes. Another difference to the inter-layer intra prediction mode in the SVC extension of H.264/AVC is that the inter-layer intra prediction mode is not only supported if the co-located area in the base layer is intra coded, but it is also supported if the co-located base layer area is inter coded or partly inter coded.
In a particular embodiment, spatial intra prediction of a difference signal (see aspect A) is performed. The multiple methods include the following method: The (potentially upsampled/filtered) reconstructed base layer signal (cp. 380) is combined with a spatial intra prediction signal, where the spatial intra prediction (cp. 420) is derived (cp. 260) based on difference samples for neighboring blocks. The difference samples represent the difference of the reconstructed enhancement layer signal (cp. 400) and the (potentially upsampled/filtered) reconstructed base layer signal (cp. 380).
The second portion of the prediction signal is generated using the co-located area 28 in the reconstructed signal 200 of the base layer. For quality enhancement layers, the co-located base layer samples can be directly used or they can be optionally filtered, e.g., by a low-pass filter or a filter 500 that attenuates high-frequency components. For spatial enhancement layers, the co-located base layer samples are upsampled. For upsampling 220, an FIR filter or a set of FIR filters can be used. It is also possible to use IIR filters. Optionally, the reconstructed base layer samples 200 can be filtered before upsampling or the base layer prediction signal (signal obtained after upsampling the base layer) can be filtered after the upsampling stage. The reconstruction process of the base layer can include one or more additional filters such as a deblocking filter (cp. 120) and an adaptive loop filter (cp. 140). The base layer reconstruction 200 that is used for upsampling can be the reconstruction signal before any of the loop filters (cp. 200c), or it can be the reconstruction signal after the deblocking filter but before any further filter (cp. 200b), or it can be the reconstruction signal after a particular filter or the reconstruction signal after applying all filters used in the base layer decoding process (cp. 200a).
The two generated portions of the prediction signal (the spatially predicted difference signal 746 and the potentially filtered/upsampled base layer reconstruction 380) are added 732 sample by sample in order to form the final prediction signal 420.
Transferring the just outlined aspect onto the embodiment of
In a particular embodiment, intra prediction is continued with inter-layer residual prediction (see aspect B). The multiple methods for generating an intra prediction signal using base layer data include the following method: A conventional spatial intra prediction signal (derived using neighboring reconstructed enhancement layer samples) is combined with an (upsampled/filtered) base layer residual signal (inverse transform of base layer transform coefficients or difference between base layer reconstruction and base layer prediction).
The concept shown in
The base layer residual signal 480 that is used for inter-layer prediction can be the residual signal that is obtained by scaling and inverse transforming 560 the transform coefficient levels of the base layer. Or it can be the difference between the reconstructed base layer signal 200 (before or after deblocking and additional filtering or between any filtering operations) and the prediction signal 660 used in the base layer.
The two generated signal components (spatial intra prediction signal 764 and inter-layer residual prediction signal 754) are added 752 together to form the final enhancement layer intra prediction signal.
This means, the prediction mode just outlined with respect to
In a particular embodiment, a weighted prediction of spatial intra prediction and base layer reconstruction (see aspect C) is used. This actually does represent the above announced specification of a specific implementation of the embodiment outlined above with respect to
The multiple methods for generating an intra prediction signal using base layer data include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with a spatial intra prediction signal, where the spatial intra prediction is derived based on reconstructed enhancement layer samples of neighboring blocks. The final prediction signal is obtained by weighting (cp. 41) the spatial prediction signal and the base layer prediction signal in a way that different frequency components use a different weighting. This can be for example realized by filtering (cp. 62) the base layer prediction signal (cp. 38) with a low-pass filter and filtering (cp. 64) the spatial intra prediction signal (cp. 34) with a high-pass filter and adding up (cp. 66) the obtained filtered signals. Or the frequency-based weighting can be realized by transforming (cp. 72, 74) the base layer prediction signal (cp. 38) and the enhancement layer prediction signal (cp. 34) and superimpose the obtained transform blocks (cp. 76, 78), where different weighting factors (cp. 82, 84) are used for different frequency positions. The obtained transform block (cp. 42 in
The concept of
When comparing reference signs used in
The two prediction signals (potentially upsampled/filtered base layer reconstruction 386 and enhancement layer intra prediction 782) are combined to form the final prediction signal 420. The method for combining these signals can have the property that different weighting factors are used for different frequency components. In a particular embodiment, the upsampled base layer reconstruction is filtered with a low-pass filter (cp. 62) (it is also possible to filter the base layer reconstruction before the upsampling 220) and the intra prediction signal (cp. 34 obtained by 30) is filtered with a high-pass filter (cp. 64), and both filtered signals are added 784 (cp. 66) to form the final prediction signal 420. The pair of low-pass and high-pass filter may represent a quadrature mirror filter pair, but this is not required.
In another particular embodiment (cp.
In one embodiment, the obtained transform coefficient block (cp. 42) (obtained by summing up the weighted transformed signals for both components) is inverse transformed (cp. 84) to form the final prediction signal 420 (cp. 54). In another embodiment, the prediction is directly done in the transform domain. That is, the coded transform coefficient levels (cp. 59) are scaled (i.e., inverse quantized) and added (cp. 52) to the transform coefficients (cp. 42) of the prediction signal (which is obtained by summing up the weighted transformed signals for both components), and then the resulting block of transform coefficients (not shown in
A selection of a base layer reconstruction and residual signal (see aspect D) may be used, too. For methods (as described above) that use the reconstructed base layer signal, the following versions can be used:
Reconstructed base layer samples 200c before deblocking 120 and further in-loop processing 140 (such as sample adaptive offset filter or adaptive loop filter).
Reconstructed base layer samples 200b after deblocking 120 but before further in-loop processing 140 (such as sample adaptive offset filter or adaptive loop filter).
Reconstructed base layer samples 200a after deblocking 120 and further in-loop processing 140 (such as sample adaptive offset filter or adaptive loop filter) or between multiple in-loop processing steps.
The selection of the corresponding base layer signal 200a,b,c can be fixed for a particular decoder (and encoder) implementation, or it can be signaled inside the bitstream 6. For the latter case, different versions can be used. The usage of a particular version of the base layer signal can be signaled at a sequence level, or at a picture level, or at a slice level, or at a largest coding unit level, or at a coding unit level, or a prediction block level, or at a transform block level, or at any other block level. In another version, the selection can be made dependent on other coding parameters (such as coding modes) or on the characteristic of the base layer signal.
In another embodiment, multiple versions of the methods that use the (upsampled/filtered) base layer signal 200 can be used. For example, two different modes that directly use the upsampled base layer signal, i.e. 200a, can be provided, where the two modes use different interpolation filters or one mode uses an additional filtering 500 of the (upsampled) base layer reconstruction signal. Similarly, multiple different versions for the other modes described above can be provided. The employed upsampled/filtered base layer signal 380 for the different versions of a mode can differ in the used interpolation filters (including interpolation filters that also filter the integer-sample positions), or the upsampled/filtered base layer signal 380 for the second version can be obtained by filtering 500 the upsampled/filtered base layer signal for the first version. The selection of one of the different version can be signaled at a sequence, picture, slice, largest coding unit, coding unit level, prediction block level, or transform block level, or it can be inferred from the characteristics of the corresponding reconstructed base layer signal or transmitted coding parameters.
The same applies to the mode that uses the reconstructed base layer residual signal, via 480. Here, also different versions can be used that differ in the used interpolation filters or additional filtering steps.
Different filters can be used for upsampling/filtering the reconstructed base layer signal and the base layer residual signal. That means that for upsampling the base layer residual signal, a different approach is used than for upsampling the base layer reconstruction signal.
For base layer blocks for which the residual signal is zero (i.e., no transform coefficient levels have been transmitted for a block), the corresponding base layer residual signal can be replaced with another signal derived from the base layer. This can be, for example, a high-pass filtered version of the reconstructed base layer block or any other difference-like signal derived from the reconstructed base layer samples or reconstructed base layer residual samples of neighboring blocks.
As far as samples used for spatial intra prediction in the enhancement layer (see aspect H) the following special treatments can be provided. For modes that use a spatial intra prediction, non-available neighboring samples in the enhancement layer (neighboring samples can be un-available, because neighboring blocks can be coded after the current block) can be replaced with the corresponding samples of an upsampled/filtered base layer signal.
As far as the coding of intra prediction modes (see aspect X) is concerned, the following special modes and functionalities may be provided. For modes that use a spatial intra prediction like 30 a, the coding of the intra prediction mode can be modified in a way that information about the intra prediction mode in the base layer (if available) is used for more efficiently coding the intra prediction mode in the enhancement layer. This could be used, for example, for the parameters 56. If the co-located area (cp. 36, in the base layer is intra coded using particular spatial intra prediction mode, it is likely that a similar intra prediction mode is also used in the enhancement layer block (cp. 28). The intra prediction mode is typically signaled in a way that among the set of possible intra prediction modes one or more modes are classified as most probable modes, which can then be signaled with shorter code words (or less arithmetic codes binary decisions results in less bits). In the intra prediction of HEVC, the intra prediction mode of the block to the top (if available) and the intra prediction mode of the block to the left (if available) are included in the set of most probable modes. In addition to these modes, one or more additional modes (which are often used) are included into the list of most probable modes, where the actual added modes depend on the availability of the intra prediction modes of the block above the current block and the block to the left of the current block. In HEVC, exactly 3 modes are classified as most probable modes. In H.264/AVC, one mode is classified as most probable mode, this mode is derived based on the intra prediction modes that are used for the block above the current block and the block to the left of the current block. Any other concept (different from H.264/AVC and HEVC) for classifying the intra prediction modes is possible, and can be used for the following extension.
In order to use the base layer data for an efficient coding of the intra prediction modes in the enhancement layer, the concept of using one or more most probable modes is modified in a way that the most probable modes include the intra prediction mode that was used in the co-located base layer block (if the corresponding base layer block was intra coded). In a particular embodiment, the following approach is used: Given the current enhancement layer block, a co-located base layer block is determined. In a particular version, the co-located base layer block is the base layer block that covers the co-located position of the top-left sample of the enhancement block. In another version, the co-located base layer block is the base layer block that covers the co-located position of a sample in the middle of the enhancement block. In other versions, other sample inside the enhancement layer block can be used for determining the co-located base layer block. If the determined co-located base layer block is intra coded and the base layer intra prediction mode specifies an angular intra prediction mode and the intra prediction mode derived from the enhancement layer block to the left of the current enhancement layer block does not use an angular intra prediction mode, then the intra prediction mode derived from the left enhancement layer block is replaced with the corresponding base layer intra prediction mode. Otherwise, if the determined co-located base layer block is intra coded and the base layer intra prediction mode specifies an angular intra prediction mode and the intra prediction mode derived from the enhancement layer block above the current enhancement layer block does not use an angular intra prediction mode, then the intra prediction mode derived from the above enhancement layer block is replaced with the corresponding base layer intra prediction mode. In other versions, a different approach for modifying the list of most probable modes (which may consist of a single element) using the base layer intra prediction mode is used.
Inter coding techniques for spatial and quality enhancement layers are presented next.
In the state-of-the-art hybrid video coding standards (such as H.264/AVC or the upcoming HEVC), the pictures of a video sequence are divided into blocks of samples. The block size can either be fixed or the coding approach can provide a hierarchical structure which allows blocks to be further subdivided into blocks with smaller block sizes. The reconstruction of a block is typically obtained by generating a prediction signal for the block and adding a transmitted residual signal. The residual signal is typically transmitted using transform coding, which means the quantization indices for transform coefficients (also referred to as transform coefficient levels) are transmitted using entropy coding techniques, and at the decoder side, these transmitted transform coefficient levels are scaled and inverse transformed to obtain the residual signal which is added to the prediction signal. The residual signal is generated either by intra prediction (using only already transmitted data for the current time instant) or by inter prediction (using already transmitted data for different time instants).
In inter prediction the prediction block is derived by motion-compensated prediction using samples of already reconstructed frames. This can be done by unidirectional prediction (using one reference picture and one set of motion parameters), or the prediction signal can be generated by multi-hypothesis prediction. In the latter case, two or more prediction signals are superimposed, i.e., for each sample, a weighted average is constructed to form the final prediction signal. The multiple prediction signals (which are superimposed) can be generated by using different motion parameters for the different hypotheses (e.g., different reference pictures or motion vectors). For uni-directional prediction, it is also possible to multiply the samples of the motion-compensated prediction signal with a constant factor and add a constant offset in order to form the final prediction signal. Such a scaling and offset correction can also be used for all or selected hypothesis in multi-hypotheses prediction.
In scalable video coding also the base layer information can be utilized to support the inter prediction process for the enhancement layer. In the state-of-the-art video coding standard for scalable coding, the SVC extension of H.264/AVC, there is one additional mode for improving the coding efficiency of the inter prediction process in an enhancement layer. This mode is signaled at a macroblock level (a block of 16×16 luma samples). In this mode, the reconstructed residual samples in the lower layer are used for improving the motion-compensated prediction signal in the enhancement layer. This mode is also referred to as inter-layer residual prediction. If this mode is selected for a macroblock in a quality enhancement layer, the inter-layer prediction signal is built by the co-located samples of the reconstructed lower layer residual signal. If the inter-layer residual prediction mode is selected in a spatial enhancement layer, the prediction signal is generated by upsampling the co-located reconstructed base layer residual signal. For upsampling, FIR filters are used, but no filtering is applied across transform block boundaries. The prediction signal that is generated from the reconstructed base layer residual samples is added to the conventional motion-compensated prediction signal to form the final prediction signal for the enhancement layer block. In general, for the inter-layer residual prediction mode, an additional residual signal is transmitted by transform coding. The transmission of the residual signal can also be omitted (inferred to be equal to zero) if it is correspondingly signaled inside the bitstream. The final reconstruction signal is obtained by adding the reconstructed residual signal (obtained by scaling the transmitted transform coefficient levels and applying an inverse spatial transform) to the prediction signal (which is obtained by adding the inter-layer residual prediction signal to the motion-compensated prediction signal).
Next, techniques for inter coding of an enhancement layer signal are described. This section describes methods for employing base layer signals in addition to already reconstructed enhancement layer signals for inter predicting the enhancement layer signal to be coded in a scalable video coding scenario. By employing the base layer signals for inter-predicting the enhancement layer signal to be coded, the prediction error can be significantly reduced, which results in an overall bit rate saving for coding the enhancement layer. The main focus of this section is to augment the block based motion compensation of enhancement layer samples using already coded enhancement layer samples with additional signals from the base layer. The following description provides possibilities to use various signals from the coded base layer. The examples presented are applicable to general block-based hybrid coding approaches without assuming any particular block partitioning, although a quad-tree block partitioning is generally employed as an embodiment. The usage of base layer reconstruction of current time index, base layer residual of current time index, or even base layer reconstruction of already coded pictures for inter prediction of enhancement layer block to be coded is described. It is also described how the base layer signals can be combined with already coded enhancement layer signals to obtain a better prediction for current enhancement layer.
One of the main techniques in state-of-the-art is the inter-layer residual prediction in H.264/SVC. Inter-layer residual prediction in H.264/SVC can be employed for all inter-coded macroblocks regardless whether they are coded using the SVC macroblock type signalled by the base mode flag or by using any of the conventional macroblock types. A flag is added to the macroblock syntax for spatial and quality enhancement layers, which signals the usage of inter-layer residual prediction. When this residual prediction flag is equal to 1, the residual signal of the corresponding area in the reference layer is block-wise upsampled using a bilinear filter and used as prediction for the residual signal of the enhancement layer macroblock, so that only the corresponding difference signal needs to be coded in the enhancement layer.
For the description in this section, the following notation is used:
-
- T0:=time index of the current picture
- T1=time index of an already reconstructed picture
- EL:=enhancement layer
- BL:=base layer
- EL(t0):=current enhancement layer picture to be coded
- EL_reco:=enhancement layer reconstruction
- BL_reco:=base layer reconstruction
- BL_resi:=base layer residual signal (inverse transform of base layer transform coefficients or difference between base layer reconstruction and base layer prediction)
- EL_diff:=difference between enhancement layer reconstruction and upsampled/filtered base layer reconstruction
The different base layer and enhancement layer signals that are used in the description are illustrated in
For the description, the following properties of filters are used:
-
- Linearity: Most filters referred to in the description are linear, but also non-linear filters may be used.
- Number of output samples: In the upsampling operation, the number of output samples is larger than the number of input samples. Here the filtering of the input data produces more samples than input values. In conventional filtering, the number of output samples is equal to the number of input samples. Such filtering operations can, for example, used in quality scalable coding.
- Phase delay: For the filtering of samples at integer positions, the phase delay is typically zero (or an integer-valued delay in samples). For generating samples of fractional positions (for example at half-pel or quarter-pel positions), typically filters with a fractional delay (in units of samples) are applied to the samples of the integer grid.
Conventional motion-compensated prediction as used in all hybrid video coding standards (e.g., MPEG-2, H.264/AVC, or the upcoming HEVC standard) is illustrated in
The following description applies to scalable coding with quality enhancement layers (the enhancement layer has the same resolution as the base layer, but represents the input video with a higher quality or fidelity) and scalable coding with spatial enhancement layers (the enhancement layer has a higher resolution, i.e., a larger number of samples, than the base layer). For quality enhancement layers, no upsampling of base layer signals is necessitated, but a filtering of the reconstructed base layer samples may be applied. In case of a spatial enhancement layers, an upsampling of the base layer signals is generally necessitated.
The embodiments support different methods for using the reconstructed base layer samples or base layer residual samples for inter prediction of an enhancement layer block. It is possible to support one or more of the methods described below in addition conventional inter prediction and intra prediction. The usage of a particular method can be signaled at the level of the largest supported block size (such as a macroblock in H.264/AVC or a coding tree block/largest coding unit in HEVC), or it can be signaled at all supported block sizes, or it can be signaled for a subset of the supported block sizes.
For all methods described in the following, the prediction signal can be directly used as reconstruction signal for the block. Or the selected method for inter-layer inter prediction can be combined with residual coding. In a particular embodiment the residual signal is transmitted via transform coding, i.e., quantized transform coefficients (transform coefficient levels) are transmitted using an entropy coding technique (e.g., variable length coding or arithmetic coding) and the residual is obtained by inverse quantizing (scaling) the transmitted transform coefficient levels and applying an inverse transform. In a particular version, the complete residual block that corresponds to the block for which the inter-layer inter prediction signal is generated is transformed using a single transform (i.e., the entire block is transformed using a single transform of the same size as the prediction block). In another embodiment, the prediction block can be further subdivided (e.g., using a hierarchical decomposition) into smaller block and for each of the smaller block (which can also have different block sizes) a separate transform is applied. In a further embodiment, a coding unit can be divided into smaller prediction blocks and for zero or more of the prediction blocks the prediction signal is generated using one of the methods for inter-layer inter prediction. And then, the residual of the entire coding unit is transformed using a single transform, or the coding unit is subdivided into different transform units, where the subdivision for forming the transform units (blocks to which a single transform is applied) is different from the subdivision for decomposing a coding unit into prediction blocks.
In the following, possibilities of performing the prediction using the base layer residual and the enhancement layer reconstruction are described. The multiple methods include the following method: A conventional inter prediction signal (derived by motion-compensated interpolation of already reconstructed enhancement layer pictures) is combined with an (upsampled/filtered) base layer residual signal (inverse transform of base layer transform coefficients or difference between base layer reconstruction and base layer prediction). This method is also referred to as the BL_resi mode (cp.
In short, the prediction for enhancement layer samples can be written as:
EL prediction=filter(BL_resi(t0))+MCP_filter(EL_reco(t1)).
It is also possible that 2- or more hypotheses of enhancement layer reconstruction signals are used, for example,
EL prediction=filter(BL_rest(t0))+MCP_filter1(EL_reco(t1))+MCP_filter2(EL_reco(t2)).
The motion-compensated prediction (MCP) filters used on enhancement layer (EL) reference pictures can be of integer or fractional sample accuracy. The MCP filters used on the EL reference pictures can be the same or different from the MCP filters used on the BL reference pictures during the BL decoding process.
A motion vector MV(x,y,t) is defined to point to a particular location in an EL reference picture. The parameters x and y point to a spatial location within a picture and the parameter t is used to address the time index of reference pictures, also called as reference index. Often the term motion vector is used to refer only to the two spatial components (x,y). The integer part of a MV is used to fetch a set of samples from a reference picture and the fractional part of a MV is used to select a MCP filter from a set of filters. The fetched reference samples are filtered to produce filtered reference samples.
The motion vectors are generally coded using differential prediction. That means, a motion vector predictor is derived based on already coded motion vectors (and potentially a syntax element indicating the used of one of a set of potential motion vector predictors), and a difference vector is included in the bitstream. The final motion vector is obtained by adding the transmitted motion vector difference to the motion vector predictor. Typically, it is also possible to completely derive the motion parameters for a block. Therefore, typically, a list of potential motion parameter candidates is constructed based on already coded data. This list can include the motion parameters of spatially neighboring blocks as well as motion parameters that are derived based on motion parameters of a co-located block in a reference frame.
The base layer (BL) residual signal can be defined as one of the following:
-
- inverse transform of BL transform coefficients, or
- difference between BL reconstruction and BL prediction, or
- for BL blocks for which the inverse transform of BL transform coefficients is zero, it can be replaced with another signal derived from the BL, e.g., a high-pass filtered version of the reconstructed BL block, or
- a combination of the above methods.
For computing the EL prediction component from the current BL residual, the region in BL picture co-located with the considered region in EL picture is identified and the residual signal is taken from the identified BL region. A definition of co-located region can be made such that it accounts for integer scaling factor of BL resolution (e.g. 2× scalability), or fractional scaling factor of BL resolution (e.g. 1.5× scalability) or even producing the same EL resolution as the BL resolution (e.g. quality scalability). In case of quality scalability, the co-located block in the BL picture has the same coordinates as the EL block to be predicted.
The co-located BL residual can be upsampled/filtered to generate filtered BL residual samples.
The final EL prediction is obtained by adding the filtered EL reconstruction samples and the filtered BL residual samples.
The multiple methods concerning prediction using a base layer reconstruction and enhancement layer difference signal (see aspect J) include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with a motion-compensated prediction signal, where the motion-compensated prediction signal is obtained by motion compensating difference pictures. The difference pictures represent the difference of the reconstructed enhancement layer signal and the (upsampled/filtered) reconstructed base layer signal for the reference pictures. This method is also referred to as the BL_reco mode.
This concept is illustrated in
EL prediction=filter(BL_reco(t0))+MCP_filter(EL_diff(t1)).
It is also possible that 2- or more hypotheses of EL difference signals are used, e.g.,
EL prediction=filter(BL_resi(t0))+MCP_filter1(EL_diff(t1))+MCP_filter2(EL_diff(t2)).
For the EL difference signal, the following versions can be used:
-
- difference between EL reconstruction and upsampled/filtered BL reconstruction, or
- difference between EL reconstruction before or between loop filtering stages (like Deblocking, SAO, ALF) and upsampled/filtered BL reconstruction.
The usage of a particular version can be can be fixed in the decoder or it can be signaled at a sequence level, picture level, slice level, largest coding unit level, coding unit level, or other partitioning levels. Or it can be made dependent on other coding parameters.
When the EL difference signal is defined to use the difference between EL reconstruction and upsampled/filtered BL reconstruction, it becomes amenable to just save EL reconstruction and BL reconstruction and compute the EL difference signal on-the-fly for the blocks using this prediction mode, thereby saving the memory necessitated to store the EL difference signal. However, it may incur a slight computational complexity overhead.
The MCP filters used on EL difference pictures can be of integer or fractional sample accuracy.
-
- For MCP of difference pictures, different interpolation filters than for MCP of reconstructed pictures can be used.
- For MCP of difference pictures, interpolation filters can be selected based on the characteristics of the corresponding area in the difference picture (or based on coding parameters or based on information transmitted in the bitstream).
A motion vector MV(x,y,t) is defined to point to a particular location in an EL difference picture. The parameters x and y point to a spatial location within a picture and the parameter t is used to address the time index of the difference picture.
The integer part of a MV is used to fetch a set of samples from a difference picture and the fractional part of a MV is used to select a MCP filter from a set of filters. The fetched difference samples are filtered to produce filtered difference samples.
The dynamic range of difference pictures can theoretically exceed the dynamic range of original pictures. Assuming an 8-bit representation of images in the range O 255, the difference images can have a range of [−255 255]. However, in practice, most of the amplitudes are distributed around the positive and negative vicinity of 0. In an embodiment of storing the difference images, a constant offset of 128 is added and the result is clipped to the range [0 255] and stored as regular 8-bit images. Later, in the encoding and decoding process, the offset of 128 is subtracted back from the difference amplitude loaded from the difference pictures.
For methods that use the reconstructed BL signal, the following versions can be used. This can be fixed or it can be signaled at a sequence level, picture level, slice level, largest coding unit level, coding unit level, or other partitioning levels. Or it can be made dependent on other coding parameters.
-
- Reconstructed base layer samples before deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking but before further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter) or between multiple in-loop processing steps.
For computing the EL prediction component from current BL reconstruction, the region in BL picture co-located with the considered region in EL picture is identified and the reconstruction signal is taken from the identified BL region. A definition of co-located region can be made such that it accounts for integer scaling factor of BL resolution (e.g. 2× scalability), or fractional scaling factor of BL resolution (e.g. 1.5× scalability) or even producing the same EL resolution as the BL resolution (e.g. SNR scalability). In case of SNR scalability, the co-located block in the BL picture has the same coordinates as the EL block to be predicted.
The final EL prediction is obtained by adding the filtered EL difference samples and the filtered BL reconstruction samples.
Some possible variations of the mode that combines an (upsampled/filtered) base layer reconstruction signal and a motion-compensated enhancement layer difference signal are listed below:
-
- Multiple versions of the methods that use the (upsampled/filtered) BL signal can be used. The employed upsampled/filtered BL signal for these versions can differ in the used interpolation filters (including interpolation filters that also filter the integer-sample positions), or the upsampled/filtered BL signal for the second version can be obtained by filtering the upsampled/filtered BL signal for the first version. The selection of one of the different version can be signaled at a sequence, picture, slice, largest coding unit, coding unit level, or other levels of picture partitioning, or it can be inferred from the characteristics of the corresponding reconstructed BL signal or the transmitted coding parameters.
- Different filters can be used for upsampling/filtering the BL reconstructed signal in case of the BL_reco mode and the BL residual signal in case of the BL_resi mode.
- It is also possible that the upsampled/filtered BL signal is combined with 2- or more hypotheses of motion-compensated difference signals. This is illustrated in
FIG. 32
Considering the above, a prediction may be performed by using a combination of base layer reconstruction and enhancement layer reconstruction (see aspect C). One major difference to the description above with respect to
-
- Filtering the base layer prediction signal with a low-pass filter and filtering the inter prediction signal with a high-pass filter and adding up the obtained filtered signals.
- Transform the base layer prediction signal and the inter prediction signal and superimpose the obtained transform blocks, where different weighting factors are used for different frequency positions. The obtained transform block can then be inverse transformed and used as enhancement layer prediction signal or the obtained transform coefficients can be added to the scaled transmitted transform coefficient levels and then inverse transformed in order to obtain the reconstructed block before deblocking and in-loop processing.
This mode is may also be referred to as BL_comb mode illustrated in
In short, the EL prediction can be expressed as,
EL prediction=BL_weighting(BL_reco(t0))+EL weighting(MCP_filter(EL_reco(t1))
In an embodiment, the weighting is made dependent on the ratio of EL resolution to the BL resolution. For instance, when the BL is to be scaled up by a factor in the range 1, 1.25) certain set of weights for the EL and BL reconstruction can be used. When the BL is to be scaled up by a factor in the range [1.25 1.75], a different set of weights can be used. When the BL is to be scaled up by a factor of 1.75 or above, a further different set of weights can be used, and so on.
Rendering the specific weighting dependent on the scaling factor separating base and enhancement layer is also feasible with the other embodiment concerning spatial intra-layer prediction.
In another embodiment, the weighting is made dependent on the EL block size to be predicted. For instance, for a 4×4 block in the EL, a weighting matrix can be defined that specifies the weighting for the EL reconstruction transform coefficients and another weighting matrix can be defined that specifies the weighting for the BL reconstruction transform coefficients. A weighting matrix for BL reconstruction transform coefficients can be, e.g.,
64, 63, 61, 49,
63, 62, 57, 40,
61, 56, 44, 28,
49, 46, 32, 15,
and a weighting matrix for the EL reconstruction transform coefficients can be, e.g.,
0, 2, 8, 24,
3, 7, 16, 32,
9, 18, 20, 26,
22, 31, 30, 23,
Similarly for block sizes like 8×8, 16×16, 32×32 and others, separate weighting matrices can be defined.
The actual transform used for the frequency domain weighting can be the same or different from the transform used to code the prediction residual. For example, an integer approximation for the OCT can be used for both frequency domain weighting and to compute transform coefficients of prediction residual to be coded in frequency domain.
In another embodiment, a maximum transform size is defined for the frequency domain weighting, in order to restrict the computational complexity. If the considered EL block size is greater than the maximum transform size, the EL reconstruction and BL reconstruction is spatially split into a series of adjacent sub-blocks and the frequency domain weighting is performed on the sub-blocks and the final prediction signal is formed by assembling the weighted results.
Furthermore, the weighting can be performed on luminance and chrominance components or a selected subset of the color components.
In the following, different possibilities for deriving enhancement layer coding parameters are described. The coding (or prediction)-parameters to be used for reconstructing an enhancement layer block can be derived by multiple methods from the co-located coding parameters in the base layer. The base and enhancement layers can have different spatial resolution or they can have the same spatial resolution.
In the scalable video extension of the H.264/AVC inter-layer motion prediction is performed for macroblock types, which are signaled by the syntax element base mode flag. If the base mode flag is equal to 1 and the corresponding reference macroblock in the base layer is inter-coded the enhancement layer macroblock is also inter-coded, and all motion parameters are inferred from the co-located base layer block(s). Otherwise (base mode flag is equal to 0), for each motion vector a so-called motion prediction flag syntax element is transmitted and specified whether base layer motion vectors are used as motion vector predictor. If the motion prediction flag equals 1 the motion vector predictor of the collocated reference block of the base layer is scaled according to the resolution ratio and used as the motion vector predictor. If the motion prediction flag equals to 0 the motion vector predictor is computed as specified in H.264/AVC.
In the following, methods for deriving enhancement layer coding parameters are described. The sample army associated with a base layer picture is decomposed into blocks and each block has associated coding (or prediction)-parameters. In other words, all sample positions inside a particular block have identical associated coding (or prediction)-parameters. The coding parameters may include parameters for motion compensated prediction comprised of the number of motion hypotheses, reference indices, motion vectors, motion vector predictor identifiers and merge identifiers. The coding parameters may also include intra prediction parameters such as intra prediction directions.
It can be signaled inside the bit stream that a block in the enhancement layer is coded using co-located information from the base layer.
For example, the derivation of enhancement layer coding parameters (see aspect T) could be made as follows. For an N×M block in an enhancement layer, which is signaled to be using co-located base layer information, the coding parameters associated to the sample positions inside the block can be derived based on the coding parameters associated to the co-located sample positions in the base layer sample array.
In a particular embodiment, this process is done by the following steps:
-
- 1 Derivation of coding parameters for each sample position in the N×M enhancement layer block based on the base layer coding parameters.
- 2. Derivation of a partitioning of the N×M enhancement layer block into sub-blocks, such that all sample positions inside a particular sub-block have identical associated coding parameters.
The second step can also be omitted.
Step 1 can be performed by using a function ƒc of the enhancement layer sample position pel, giving the coding parameters c, i.e.,
c=ƒc(pel)
For example, to ensure a minimum block size m×n in the enhancement layer, the function ƒc can return the coding parameters c associated to Pbl given by the function ƒp,m×n with
The distance between two horizontally or vertically adjacent base layer sample positions is thereby equal to 1, and the top left most base layer sample and the top left most enhancement layer sample both have position p=(0,0).
As another example, the function ƒc(pel) can return the coding parameters c associated to the base layer sample position pbl which is nearest to the base layer sample position Pel.
The function ƒc(pel) can also interpolate coding parameters in case the given enhancement layer sample position has fractional components in units of the distance between base layer sample positions.
Before returning motion parameters, the function ƒ, rounds the spatial displacement components of the motion parameters to the nearest available value in the enhancement layer sampling grid.
After step 1, each enhancement layer sample can be predicted, as each sample position has associated prediction parameters after step 1. Nevertheless, in step 2 a block partitioning can be derived for the purpose of performing prediction operations on larger blocks of samples, or for the purpose of transform coding the prediction residual within the blocks of the derived partitioning.
Step 2 can be performed by grouping enhancement layer sample positions into square or rectangular blocks, each being decomposed into one of a set of allowed decompositions into sub-blocks. The square or rectangular blocks correspond to leafs in a quad tree structure in which they can exist on different levels as depicted in
The level and decomposition of each square or rectangular block can be determined by performing the following ordered steps:
-
- a) Set the highest level to the level corresponding to blocks of size N×M. Set the current level to the lowest level, that is the level at which a square or rectangular block contains a single block of the minimum block size. Go to step b).
- b) For each square or rectangular block at the current level, if an allowed decomposition of the square or rectangular block exists, such that all sample positions inside each sub-block are associated to identical coding parameters, or are associated to coding parameters with small differences (according to some difference measure), that decomposition is a candidate decomposition. Of all candidate decompositions, choose the one that decomposes the square or rectangular block into the least number of sub-blocks. If the current level is the highest level, go to step c). Otherwise, set the current level to the next higher level and go to step b).
- c) Finished
The function ƒc can be chosen in such a way that there is at least one candidate decomposition at some level in step b).
The grouping of blocks with identical coding parameters is not restricted to square blocks, but the blocks can also be summarized to rectangular blocks. Furthermore, the grouping is not restricted to a quadtree structure, it is also possible to use decomposition structures in which a block is decomposed into two rectangular blocks of the same size or two rectangular blocks of different sizes. It is also possible to use a decomposition structures that uses a quadtree decomposition up to a particular level and then uses a decomposition into two rectangular blocks. Also, any other block decomposition is possible.
In contrast to the SVC inter-layer motion parameter prediction mode, the described mode is not only supported at a macroblock level (or largest supported block size), but at arbitrary block sizes. That means, the mode cannot only be signaled for the largest supported block size, but a block of the largest supported block size (macroblock in MPEG-4, H.264 and coding tree block/largest coding unit in HEVC) can be hierarchically subdivided into smaller blocks/coding units and the usage of the inter-layer motion mode can be signaled at any supported block size (for the corresponding block). In a particular embodiment, this mode is only supported for selected block sizes. Then, the syntax element that signals the usage of this mode can be transmitted only for the corresponding block sizes, or the values of a syntax element that signals (among other coding parameters) the usage of this mode can be correspondingly restricted for other block sizes. Another difference to the inter-layer motion parameter prediction mode in the SVC extension of H.264/AVC is that the block that is coded in this mode is not completely inter coded. The block can contain intra coded sub-blocks, depending on the co-located base layer signal.
One of several methods to reconstruct an M×M enhancement layer block of samples using coding parameters derived by the above described method can be signaled inside the bit stream. Such methods to predict an enhancement layer block using derived coding parameters may include the following:
-
- Derive a prediction signal for the enhancement layer block using the derived motion parameters and the reconstructed enhancement layer reference pictures for motion compensation.
- Combination of (a) (upsampled/filtered) base layer reconstruction for current picture and (b) motion compensated signal using the derived motion parameters and enhancement layer reference picture that are generated by subtracting the (upsampled/filtered) base layer reconstruction from the reconstructed enhancement layer picture.
- Combination of (a)(upsampled/filtered) base layer residual (difference between reconstructed signal and prediction or inverse transform of coded transform coefficient values) for current picture and (b) motion compensated signal using the derived motion parameters and the reconstructed enhancement layer reference pictures.
The process for deriving the partitioning into smaller blocks for the current blocks and deriving coding parameters for the sub-blocks can classify some of the sub-blocks as intra-coded while the other sub-blocks are classified as inter-coded. For inter-coded subblock, the motion parameters are also derived from the co-located base layer blocks. But if the co-located base layer block is intra coded, the corresponding subblock in the enhancement layer may be also classified as intra coded. For the samples of such intra-coded sub-blocks, the enhancement layer signal can be predicted by using information from the base layer, as for example:
-
- An (upsampled/filtered) version of the corresponding base layer reconstruction is used as intra prediction signal.
- The derived intra prediction parameters are used for spatial intra prediction in the enhancement layer.
The following embodiments to predict an enhancement layer block using a weighted combination of prediction signals contain a method to generate a prediction signal for an enhancement layer block by combining (a) an enhancement layer internal prediction signal obtained by spatial or temporal (i.e. motion compensating) prediction using reconstructed enhancement layer samples and (b) a base layer prediction signal which is the (upsampled/filtered) base layer reconstruction for the current picture. The final prediction signal is obtained by weighting the enhancement layer internal prediction signal and the base layer prediction signal in a way that for each sample a weight according to weighting function is used.
The weighting function can for example be realized by the following method. Compare a low pass filtered version of the original enhancement layer internal prediction signal v with a low pass filtered version of the base layer reconstruction u. Derive from that comparison a weight for each sample position to be used to combine the original inter prediction signal and the (upsampled/filtered) base layer reconstruction. The weight can for example be derived by mapping the difference u−v to a weight w using a transfer function t, i.e.,
t(u−v)=w
A different weighting function can be used for different block sizes of the current block to be predicted. Also, the weighting function can be modified according to the temporal distance of the reference pictures the inter prediction hypotheses are obtained from.
In case of the enhancement layer internal prediction signal being an intra prediction signal, the weighting function can for example also be realized by using a different weight depending on the position inside the current block to be predicted.
In an embodiment, the method to derive enhancement layer coding parameters is used, and step 2 of the method is using the set of allowed decompositions of a square block as depicted in
In an embodiment, the function ƒc(pel) returns the coding parameters associated to the base layer sample position given by the function ƒp,m×n(pel) described above with m=4, n=4
In an embodiment, the function ƒc(pel) returns the following coding parameters c:
-
- First, abase layer sample position is derived as pbl=(ƒ)p,4×4(pel).
- If pbl has associated inter prediction parameters that were obtained by merging with a previously coded base layer block (or has the same motion parameters), c is equal to the motion parameters of the enhancement layer block which corresponds to the base layer block that is used for merging in the base layer (i.e., the motion parameters are copied from the corresponding enhancement layer block).
- Otherwise, c equals the coding parameters associated to pb.
Also a combination of the above embodiments is possible.
In another embodiment, for an enhancement layer block which is signaled to be using co-located base layer information, associate to those enhancement layer sample positions with derived intra prediction parameters a default set of motion parameters, so that a block can be merged with the block containing these samples (i.e. copy the default set of motion parameters). The default set of motion parameters is comprised of an indicator to use one or two hypotheses, reference indices referencing the first picture in a reference picture list, and motion vectors with zero spatial displacement.
In another embodiment, for an enhancement layer block which is signaled to be using co-located base layer information, the enhancement layer samples with derived motion parameters are predicted and reconstructed first in some order. After that, the samples with derived intra prediction parameters are predicted in an intra reconstruction order. Thereby, the intra prediction can then use the already reconstructed sample values from (a) any neighboring inter prediction block and (b) neighboring intra prediction blocks which are predecessors in the intra reconstruction order.
In another embodiment, for enhancement layer blocks being merged (i.e. taking the motion parameters derived from other inter predicted blocks), the list of merge candidates additionally contains a candidate from the corresponding base layer block and if the enhancement layer has a higher spatial sampling rate than the base layer, additionally contains up to four candidates derived from the base layer candidate by refining the spatial displacement components to neighboring values only available in the enhancement layer.
In another embodiment, the difference measure used in step 2 b) states that there are small differences in a sub-block only if there are no differences at all, i.e. a sub-block can only be formed when all contained sample positions have identical derived coding parameters.
In another embodiment, the difference measure used in step 2 b) states that there are small differences in a sub-block if either (a) all contained sample positions have derived motion parameters and no pair of sample positions inside the block have derived motion parameters differing more than a particular value according to a vector norm applied to the corresponding motion vectors or (b) all contained sample positions have derived intra prediction parameters and no pair of sample positions inside the block have derived intra prediction parameters differing more than a particular angle of directional intra prediction. The resulting parameters for the sub-block are calculated by mean or median operations.
In another embodiment, the partitioning obtained by inferring coding parameters from the base layer can be further refined based on side information signaled inside the bitstream.
In another embodiment, the residual coding for a block for which the coding parameters are inferred from the base layer is independent of the partitioning into blocks that is inferred from the base layer. That means, for example, that a single transform may be applied to a block although the inferring of coding parameters from the base layer partitions the blocks into several subblock each with a separate set of coding parameters. Or the block for which the partitioning and the coding parameters for the subblocks are inferred from the base layer may be split into smaller blocks for the purpose of transform coding the residual, where the splitting into transform blocks is independent from the inferred partitioning into blocks with different coding parameters.
In another embodiment, the residual coding for a block for which the coding parameters are inferred from the base layer is dependent of the partitioning into blocks that is inferred from the base layer. That means, for example, that for transform coding the splitting of the block in transform blocks is dependent of the partitioning that is inferred from the base layer. In one version, a single transform may be applied to each of the subblocks with different coding parameters. In another version, the partitioning may be refined based on side information include into the bitstream. In another version, some of the subblocks may be summarized to a larger blocks as signaled inside the bitstream for the purpose of transform coding the residual signal.
Embodiments obtained by combinations of the above described embodiments are also possible.
Relating to enhancement layer motion vector coding, this following part describes a method for reducing motion information in scalable video coding applications by providing multiple enhancement layers predictors and employing the motion information coded in the base layer to code the motion information of the enhancement layer efficiently. This idea is applicable for scalable video coding including spatial, temporal and quality scalability.
In the scalable video extension of the H.264/AVC inter-layer motion prediction is performed for macroblock types, which are signaled by the syntax element base mode flag.
If the base mode flag is equal to 1 and the corresponding reference macroblock in the base layer is inter-coded the enhancement layer macroblock is also inter-coded, and all motion parameters are inferred from the co-located base layer block(s). Otherwise (base mode flag is equal to 0), for each motion vector a so-called motion prediction flag syntax element is transmitted and specified whether base layer motion vectors are used as motion vector predictor. If the motion prediction flag equals 1 the motion vector predictor of the collocated reference block of the base layer is scaled according to the resolution ratio and used as the motion vector predictor. If the motion prediction flag equals to O the motion vector predictor is computed as specified in H.264/AVC.
In HEVC motion parameters are predicted by applying advanced motion vector competition (AMVP). AMVP features two spatial and one temporal motion vector predictors that compete with each other. The spatial candidates are selected from positions of neighboring prediction blocks located to the left or above the current prediction block. The temporal candidate is selected among co-located positions of a previously coded picture. The positions of all spatial and temporal candidates are displayed in
After the spatial and temporal candidates are inferred a redundancy check is performed which may introduce zero motion vector as candidates to the list. An index addressing the candidate list is transmitted to identify the motion vector predictor that is used together with the motion vector difference for motion compensated prediction.
HEVC further employs a block merging algorithm, which targets the reduction of coding redundant motion parameters that result from quad-tree based coding schemes. This is achieved by creating regions consisting of multiple prediction blocks sharing identical motion parameters. These motion parameters only need to be coded once for the first prediction block of each region—seeding new motion information. Similar to AMVP the block merging algorithm constructs a list containing possible merge candidates for each prediction block. The number of candidates is defined by NumMergeCands, which is signaled in the slice header and ranges from 1 up to 5. The candidates are inferred from spatially neighboring prediction blocks and from prediction blocks in collocated temporal pictures. Possible sample positions for the prediction blocks that are considered as candidates are equal to the positions shown in
In cases where the number of spatial and temporal candidates is less than NumMergeCands additional candidates are provided by combining existing candidates or by inserting zero motion vector candidates. If a candidate is added to the list it is equipped with an index, which is used to identify the candidate. With the addition of a new candidate to the list the index is increased (starting from 0) until the list is completed with the last candidate identified by index NumMergeCands−1. A fixed length codeword is used for coding the merge candidate index to ensure an independent operation of the derivation of the candidate list and the parsing of the bit stream.
The following section describes a method for using multiple enhancement layer predictors including predictors derived from the base layer to code the motion parameters of the enhancement layer. Motion information that is already coded for the base layer can be employed to significantly reduce the motion data rate while coding the enhancement layer. This method includes the possibility to directly derive all motion data of a prediction block from the base layer in which case no additional motion data needs to be coded. In the following description the term prediction block refers to a prediction unit in HEVC, a M×N block in H.264/AVC and can be understood as a general set of samples in a picture.
The first part of the present section is about extending the list of motion vector prediction candidates by a base layer motion vector predictor (see aspect K). A base layer motion vector is added to the motion vector predictor list during the enhancement layer coding. This is achieved by inferring one or multiple motion vector predictors of the co-located prediction block from the base layer and use them as candidates in the list of the predictors for the motion compensated prediction. The co-located prediction block of the base layer is located at the center, to the left, above, to the right or underneath of the current block. If the prediction block of the base layer of the selected position does not contain any motion related data or resides outside of the current scope and is therefore currently not accessible alternative positions can be used to infer the motion vector predictors. These alternative positions are depicted in
The motion vectors inferred of the base layer may be scaled according to the resolution ratio before they can be used as predictor candidates. An index addressing the candidate list of motion vector predictors as well as a motion vector difference are transmitted for a prediction block, which specify the final motion vector used for motion-compensated prediction. In contrast to the scalable extension of the H.264/AVC standard the embodiments presented here do not constitute the usage of the motion vector predictor of the collocated block in the reference picture—rather it is available in a list among other predictors and can be addressed by an index which is transmitted.
In an embodiment the motion vector is derived from center position C1 of the collocated prediction block of the base layer and added to the top of the candidate list as a first entry. The candidate list of motion vector predictors is extended by one item. If there is no motion data in the base layer available for the sample position C1 the list construction is untouched. In another embodiment any sequence of sample positions in the base layer may be checked for motion data. In case motion data is found, the motion vector predictor of the corresponding position is inserted to the candidate list and available for the motion compensated prediction of the enhancement layer. Furthermore the motion vector predictor derived from the base layer may be inserted to the candidate list at any other position of the list. In another embodiment the base layer motion predictor may only be inserted to the candidate list if certain constraints are met. These constraints include the value of the merge flag of the collocated reference block, which is necessitated to equal 0. Another constraint may be the dimensions of the prediction block in the enhancement layer equaling the dimensions of the collocated prediction block of the base with respect to the resolution ratio. For example in the application of K×spatial scalability—if the width of the collocated block in the base layer equals to N the motion vector predictor may only be inferred if the width of the prediction block to be coded in the enhancement layer equals to K*N.
In another embodiment more than one motion vector predictors from several sample position of the base layer may be added to the candidate list of the enhancement layer. In another embodiment the candidate with the motion vector predictor inferred from the collocated block may replace spatial or temporal candidates in the list rather than extending the list. It is also possible to include multiple motion vector predictors derived from base layer data into the motion vector predictor candidate list.
The second part is about extending the list of merge candidates by a base layer candidate (see aspect K). Motion data of one or more collocated blocks of the base layer is added to the merge candidate list. This method enables the possibility to create a merge region sharing the identical motion parameters across the base and the enhancement layer. Similar to the previous section the base layer block that covers the collocated sample at the center position is not limited to this center position but can be derived from any position in the immediate vicinity, as depicted in
In an embodiment the motion vector predictor of the collocated block in the base layer covering the sample position C1 of
The third part of this specification is about reordering the motion parameter (or merge) candidate list using base layer data (see aspect L) and describes the process of re-ordering the merge candidate list according to the information already coded in the base layer. If the collocated base layer block that covers the sample of the current block is motion compensated predicted with a candidate derived from a particular origin, the corresponding enhancement layer candidate from the equivalent origin (if existent) is put as the first entry at the top of the merge candidate list. This step is equal to addressing this candidate with the lowest index, which results in assigning the cheapest codeword to this candidate.
In an embodiment the collocated base layer block is motion compensated predicted with a candidate that originates from a prediction block that covers the sample position A1 as depicted in
The fourth part of this specification is about reordering the motion vector predictor candidate list using base layer data (see aspect L) and embodies the process of re-ordering the candidate list of the motion vector prediction using the motion parameters of the base layer block. If the collocated base layer block that covers the sample of the current prediction block uses a motion vector from a particular origin, the motion vector predictor from the corresponding origin in the enhancement layer is used as a first entry in the motion vector predictor list of the current prediction block. This results in assigning the cheapest codeword to this candidate.
In an embodiment the collocated base layer block is motion compensated predicted with a candidate that originates from a prediction block that covers the sample position A1 as depicted in
The following is about enhancement layer coding of transform coefficients.
In the state-of-the-art video and image coding the residual of a prediction signal is forward transformed and the resulting quantized transform coefficients are signaled within the bit stream. This coefficient coding follows a fixed scheme:
Depending on the transformation size (for luma residuals: 4×4, 8×8, 16×16 and 32×32) different scan directions are defined. Given the first and the last position in scan order, these scans uniquely determine which coefficient positions can be significant, and thus need to be coded. In all scans the first coefficient is set to be the DC coefficient at position (0,0), whereas the last position has to be signaled within the bitstream, which is done by coding its x (horizontal) and y (vertical) position within the transformation block. Starting from the last position, the signaling of significant coefficients is done in a reverse scan order until the DC position is reached.
For transformation sizes 16×16 and 32×32 only one scan is defined, i.e. the ‘diagonal scan’, whereas transformation blocks of sizes 2×2, 4×4, and 8×8 can additionally make use of the ‘vertical’ and ‘horizontal’ scan. However, the use of the vertical and horizontal scan is restricted to residuals of intra predicted coding units and the actual used scan is derived from the direction mode of that intra prediction. Direction modes with an index in the range of 6 and 14 result in a vertical scan, whereas direction modes with an index in the range of 22 and 30 result in a horizontal scan. All remaining direction modes result in a diagonal scan.
In zigzag scan, for blocks of size greater than 4×4 the subgroup consists of 4×4 pixels block scanned in a zigzag scan. The subgroups are scanned in a zigzag manner.
The following section describes extensions for transformation coefficient coding. These include an introduction of new scan modes, methods of assigning scans to transformation blocks and a modified coding of significant coefficient positions. These extensions allow a better adaptation to different coefficient distributions within the transformation block, and thus achieve a coding gain in rate-distortion sense.
New realizations for vertical and horizontal scan patterns are introduced for 16×16 and 32×32 transformation blocks. In contrast to previously proposed scan patterns, the size of a scan subgroup is, 16×1 for horizontal scan, and 1×16 for vertical scans, respectively. Subgroups with a size of 8×2 and 2×8, respectively, might be also chosen. The subgroups themselves are scanned in the same manner.
The vertical scan is efficient for transformed coefficients that are located in a column wise spread. This can be found in images that contain horizontal edges.
The horizontal scan is efficient for transformed coefficients that are found in a row wise spread. This can be found in images that contain vertical edges.
VerHor scan is a introduced scan pattern, which allows coding of coefficients in a column by row wise scan. For 4×4 blocks the first column is scanned followed by the rest of the first row then the rest of the second column then the rest of the coefficients in second row. Then the rest of the third column is scanned and finally the rest of the fourth row and column.
For larger blocks the block is divided into 4×4 subgroups. These 4×4 blocks are scanned in a VerHor scan, whereas the subgroups are scanned VerHor scan themselves.
The vector scan can be used in cases where coefficients are located in the first columns and rows in the block. In this way, the coefficients are scanned earlier than the cases when using other scans for example diagonal scan. This can be found for images that contain both horizontal and vertical edges.
Other Scans are feasible as well. All combinations between the scans and the subgroups can be used, for example. For example using the horizontal scan for the 4×4 blocks with the diagonal scan of the subgroups. Adaptive selection of scans can be applied by selecting a different scan for each subgroup.
It should be mentioned that the different scans can be realized in a way that the transform coefficients are re-ordered after quantization at the encoder side and the conventional coding is used. At the decoder side, the transform coefficients are conventionally decoded and re-ordered before the scaling and inverse transform (or after the scaling and before the inverse transform).
Different parts of the base layer signal might be utilized to derive coding parameters from baselayer signal. Among those signals are:
-
- co-located reconstructed base layer signal
- co-located residual base layer signal
- estimated enhance layer residual signal, obtained by subtracting the enhancement layer prediction signal from the reconstructed base layer signal
- picture partitioning of the base layer frame.
Gradient parameters may be derived as follows:
For each pixel of the investigated block, a gradient is calculated. From these gradients the magnitude and angle are calculated. The angle that occurred most in the block is associated to the block (block angle). The angles are rounded so that only three directions are used horizontal (0°), vertical (90°) and diagonal (45°).
Detecting Edges:An edge detector may be applied on the investigated block as follows:
First the block is smoothed by n x n smoothing filter (e.g. Gaussian).
A gradient matrix of size m×m is used to calculate the gradient on each pixel. The magnitude and angle of every pixel is calculated. The angles are rounded so that only three directions are used horizontal (0°), vertical (90°) and diagonal (45°).
For every pixel that has a magnitude greater than a certain threshold1, the neighboring pixels are checked. If the neighboring pixel has a magnitude higher than a threshold2 and has the same angle as the current pixel then the counter of this angle is increased. For the whole block the counter with the highest number is selected as the angle of the block.
Obtaining Base Layer Coefficients by Forward TransformationIn order to derive coding parameters, for a particular TU, from the frequency domain of the base layer signal, the investigated co-located signal (reconstructed base layer signal/residual base layer signal/estimated enhancement layer signal) might be transformed into frequency domain. Advantageously, this is done using the same transformation, as is used by that particular enhancement layer TU.
The resulting base layer transformation coefficients might be quantized, or not.
In order to obtain comparable coefficient distributions as in the enhancement layer block, rate distortion quantization with a modified lambda could be used.
Scan Effectiveness Score of a Given Distribution and ScanA scan effectiveness score of a given significant coefficient distribution might be defined as follows:
Let each position of an investigated block be represented by its index in order of the investigated scan. Then, the sum of index values of significant coefficients positions is defined as the effectiveness score of this scan. Thus, scans with a smaller score, represent the particular distribution more efficiently.
Adaptive Scan Pattern Selection for Transformation Coefficient CodingIf several scans are available for a particular TU, a rule needs to be defined which uniquely selects one of them.
Methods for Scan Pattern SelectionThe selected scan can be directly derived from already decoded signals (without transmitted any additional data). This can be done, either based on the properties of the co-located base layer signal, or by utilizing enhancement layer signals only.
The scan pattern can be derived from the EL signal by
-
- State-of-the art derivation rule as described above.
- Using that scan pattern for chrominance residuals, that was chosen for the co-located luminance residual
- Defining a fixed mapping between coding modes and used scan pattern.
- Derivation the scan pattern from the last significant coefficient position (relative to an assumed fixed scan pattern).
In an embodiment, the scan pattern is selected depending on the already decoded last position as follows:
The last position is represented as x and y coordinates within the transformation block, and already decoded (For scan dependent last coding, a fixed scan pattern is assumed for the decoding process of the last position, which can be the state-of-the-art scan pattern of that TU). Let T be a defined threshold, which might depend on the particular transformation size. If neither the x coordinate, nor the y coordinate of the last significant position exceeds T, diagonal scan is selected.
Otherwise, x is compared toy. If x exceeds y horizontal scan is chosen, vertical scan otherwise. An advantageous value of T for 4×4 TUs is 1. An advantageous value of T for TUs larger than 4×4 is 4.
In a further embodiment, the derivation of the scan pattern as described in the previous embodiment is restricted to be done only for TUs of size 16×16 and 32×32. It can be further restricted to luminance signals only.
The scan pattern can also be derived from the BL signal. For deriving the selected scan pattern from the base layer signal, any coding parameter described above can be used. In particular, a gradient of a co-located base layer signal can be calculated and compared to predefined thresholds and/or potentially found edges can be utilized.
In an embodiment, the scan direction is derived depending on the block gradient angle, as follows: For gradients quantized in horizontal direction, a vertical scan is used. For gradients quantized in vertical direction, a horizontal scan is used. Otherwise, a diagonal scan is selected.
In a further embodiment, the scan pattern is derived as described in the previous embodiment, but for those transformation blocks only, for which the number of occurrence of the block angle exceeds a threshold. The remaining transformation units are decoded using the state-of-the-art scan pattern of the TU.
If base layer coefficients of the co-located block are available, either explicitly signaled in the base layer data stream or calculated by a forward transformation, these can be utilized in the following ways:
-
- For each available scan the costs for coding the base layer coefficients can be estimated. The scan with minimum costs is used for decoding the enhancement layer coefficients.
- An effectiveness score of each available scan is calculated for the base layer coefficient distribution, the scan with the minimum score is used for decoding the enhancement layer coefficients.
- The distribution of base layer coefficients within the transformation block is classified to one of a predefined set of distributions, which is associated with a particular scan pattern.
- The scan pattern is selected depending on the last significant base layer coefficient.
If the co-located base layer block was predicted using an intra prediction, the intra direction of that prediction can be used to derive an enhancement layer scan pattern.
Furthermore, the transformation size of the co-located base layer block might be utilized for deriving the scan pattern.
In an embodiment, the scan pattern is derived from the BL signal for TUs only, which represent residuals of INTRA_COPY mode predicted blocks, and their co-located base layer block has been intra predicted. For those blocks, a modified state-of-the-art scan selection is used. In contrast to the state-of-the-art scan-selection, the intra prediction direction of the co-located base layer block is used for selecting the scan pattern.
Signaling of an Scan Pattern Index within the Bitstream (See Aspect R)
Scan patterns of transformation blocks can also be chosen by the encoder in rate-distortion sense and then signaled within the bitstream.
A specific scan pattern can be coded, by signaling an index into a list of available scan pattern candidates. This list can be either a fixed list of scan patterns defined for a particular transformation size, or can be filled dynamically within the decoding process. Filling the list dynamically, allows the adaptive picking of those scan patterns, which most likely code the particular coefficient distribution most efficiently. By doing so, the number of available scan patterns for a particular TU can be reduced, and thus, the signaling of an index into that list is less expensive. If the number of scan patterns in a particular list is reduced to one, no signaling is necessary.
The process of selecting scan pattern candidates for a particular TU may utilize any coding parameter, described above, and/or follow certain rules, which utilize specific characteristics of that particular TU. Among those are:
-
- The TU represents a residual of a luminance/chrominance signal.
- The TU has a specific size.
- The TU represents a residual of a specific prediction mode.
- The last significant position within the TU is known by the decoder and lies within a specific subdivision of the TU.
- The TU is part of an 1/B/P-Slice.
- The coefficients of the TU are quantized using specific quantization parameters.
In an embodiment, the list of scan pattern candidates contains the three scans: ‘diagonal scan’, ‘vertical scan’ and ‘horizontal scan’, for all TUs.
Further embodiments can be obtained by letting the candidate list contain an arbitrary combination of scan patterns.
In a particular embodiment, a list of scan pattern candidates may contain any of the scans: ‘diagonal scan’, ‘vertical scan’ and ‘horizontal scan’.
Whereas, the scan pattern chosen by the state-of-the-art scan derivation (as described above) is set to be first in the list. Only if a particular TU has a size of 16×16 or 32×32, further candidates are added to the list. The order of the remaining scan patterns depends on the last significant coefficient position.
(Note: diagonal scan is the first pattern in the list assuming 16×16 and 32×32 transformations)
If the magnitude of its x coordinate exceeds the y coordinates magnitude, horizontal scan is chosen next, and vertical scan is put at last position. Otherwise, vertical scan is put at 2nd position, followed by horizontal scan.
Other embodiments are obtained by further restricting the condition for having more than one candidate in the list.
In another embodiment, the vertical and horizontal scan are only added to the candidate lists of 16×16 and 32×32 transformation blocks, if their coefficients represent a residual of a luminance signal.
In another embodiment, the vertical and horizontal scan are added to the candidate lists of a transformation block, if both the x and y coordinates of the last significant position are greater than a specific threshold. This threshold can be mode and/or TU size dependent. An advantageous threshold value is 3 for all sizes larger than 4×4, and 1 for 4×4 TUs.
In another embodiment, the vertical and horizontal scan are added to the candidate lists of a transformation block, if either, the x or the y coordinates of the last significant position is greater than a specific threshold. This threshold can be mode and/or TU size dependent. An advantageous threshold value is 3 for all sizes larger than 4×4, and 1 for 4×4 TUs.
In another embodiment, the vertical and horizontal scans are only added to the candidate lists of 16×16 and 32×32 transformation blocks, if the x and y coordinates of the last significant position both are greater than a specific threshold. This threshold can be mode and/or TU size dependent. An advantageous threshold value is 3 for all sizes larger than 4×4, and 1 for 4×4 TUs.
In another embodiment, the vertical and horizontal scans are only added to the candidate lists of 16×16 and 32×32 transformation blocks, if either, the x or the y coordinates of the last significant position both is greater than a specific threshold. This threshold can be mode and/or TU size dependent. An advantageous threshold value is 3 for all sizes larger than 4×4, and 1 for 4×4 TUs.
For either of the described embodiments, where specific scan patterns are signaled within the bitstream, the signaling itself can be done at different signaling levels. In particular, the signaling can be done for each TU (which falls in the subgroup of TUs with signaled scan patterns), at any node of the residual quad-tree (all sub-TUs of that node, that use a signaled scan, use the same candidate list index), at CU/LCU level, or at slice level.
The index into the candidate list can be transmitted using fixed-length coding, variable-length coding, arithmetic coding (including context-adaptive binary arithmetic coding), or PIPE coding. If context-adaptive coding is used, the context can be derived based on the parameters of neighboring blocks, the coding modes described above and/or specific characteristics of the particular TU itself.
In an embodiment, context-adaptive coding is used for signaling an index into the scan pattern candidate list of a TU, whereas the context model is derived based on the transformation size and/or the position of the last significant position within the TU.
Either method described above for deriving scan patterns, could also be used to derive a context model for signaling an explicit scan pattern for a particular TU.
For coding the last significant scanning position, the following modifications may be used in an enhancement layer:
-
- Separate context models are used for all or a subset of the coding modes that use base layer information. It is also possible to use different context models for different modes with base layer information.
- The context modeling can depend on the data in the co-located base layer block (e.g., transform coefficient distribution in base layer, gradient information of base layer, last scanning position in co-located base layer blocks).
- The last scanning position can be coded as difference to the last base layer scanning position.
- If the last scanning position is coded by signaling its x and y positions within the TU, the context modeling of the second signaled coordinate can depend on the value of the first one.
- Either method described above for deriving scan patterns, which is independent of the last significant position, could also be used to derive context models for signaling the last significant position.
In a particular version, the scan pattern derivation depends on the last significant position:
-
- If the last scanning position is coded by signaling its x and y position within the TU, the context modeling of the second coordinate can depend on those scan patterns, which are still possible candidates, when already knowing the first coordinate.
- If the last scanning position is coded by signaling its x and y position within the TU, the context modeling of the second coordinate can depend on whether, the scan pattern is already uniquely selected, when already knowing the first coordinate.
In another version, the scan pattern derivation is independent of the last significant position:
-
- The context modeling can depend on the used scan pattern in a particular TU.
- Either method described above for deriving scan patterns, could also be used to derive context models for signaling the last significant position.
For coding significant positions within a TU and the significance flags (sub-group flags and/or significance flags for single transform coefficients), respectively, the following modifications may be used in an enhancement layer:
-
- Separate context models are used for all or a subset of the coding modes that use base layer information. It is also possible to use different context models for different modes with base layer information.
- The context modeling can depend on the data in the co-located base layer block (e.g., number of significant transform coefficients for particular frequency positions).
- Either method described above for deriving scan patterns, could also be used to derive context models for signaling significant position and/or their levels.
- A generalized template can be used that evaluated both, the number of significant already coded transform coefficient levels in a spatial neighborhood of the coefficient to be coded and the number of significant transform coefficients in the co-located base layer signal of similar frequency positions.
- A generalized template can be used that evaluated both, the number of significant already coded transform coefficient levels in a spatial neighborhood of the coefficient to be coded and the levels of significant transform coefficients in the co-located base layer signal of similar frequency positions.
- The context modeling for sub-group flags might depend on the used scan pattern and/or particular transformation sizes.
A usage of different context initialization tables for base and enhancement layer may be used. The context model initialization for the enhancement layer might be modified in the following ways:
-
- The enhancement layer uses a separate set of initialization values.
- The enhancement layer uses separate sets of initialization values, for the different operation modes (spatial/temporal or quality scalability)
- Enhancement layer context models which have counterparts in the base layer, might use the state of their counterpart as initialization states.
- The algorithm for deriving the initial states of context models, can be base layer QP and/or delta QP dependent.
Next, a possibility of backward adaptive enhancement layer coding using base layer data is described. This following part describes methods to create an enhancement layer prediction signal in a scalable video coding system. The methods use the base layer decoded picture sample information to infer the value of prediction parameters, which are not transmitted in the coded video bitstream, but which are used to form a prediction signal for the enhancement layer. Thus the overall bitrate necessitated to code the enhancement layer signal is reduced.
A state of the art hybrid video encoder decomposes the source image into blocks of different sizes, usually following a hierarchy. For each block the video signal is predicted from spatial neighboring blocks (intra prediction), or temporally previously coded pictures (inter prediction). The difference between the prediction and the actual image is transformed and quantized. The resulting prediction parameters and transform coefficients are entropy coded to form a coded video bitstream. The matching decoder follows the steps in inverse order.
Scalable video coding a bitstream is composed of different layers: a base layer offering complete decodable video and enhancement layers which can additionally be used for decoding. Enhancement layers can provide a higher spatial resolution (spatial scalability), temporal resolution (temporal scalability) or quality (SNR scalability).
In previous standards like H.264/AVC SVC the syntax elements like motion vectors, reference picture indexes or intra prediction modes are predicted directly from a corresponding syntax element in the coded base layer.
In the enhancement layer a mechanism exists at block level to switch between using a prediction signal that is derived from the base layer syntax elements or predict from other enhancement layer syntax elements or decoded enhancement layer samples.
In the following part, base layer data are used for deriving enhancement layer parameters at the decoder side.
Method 1: Motion Parameter Candidate DerivationFor a block (a) of the spatial or quality enhancement layer picture the corresponding block (b) of the base layer picture is determined, that covers the same picture area.
An inter prediction signal for block (a) of the enhancement layer is formed using the following method:
-
- 1 Motion compensation parameter set candidates are determined, e.g. from temporally or spatially neighboring enhancement layer blocks or derivatives thereof.
- 2 A motion compensation is performed for each candidate motion compensation parameter set to form an inter prediction signal in the enhancement layer.
- 3. The best motion compensation parameter set is selected by minimizing an error measure between the prediction signal for the enhancement layer block (a) and the reconstruction signal of base layer block (b). For spatial scalability the base layer block (b) can be spatially upsampled using an interpolation filer.
Motion compensation parameters sets contain a specific combination of motion compensation parameters.
Motion compensation parameters can be motion vectors, reference picture indexes, the selection between uni- and bi-prediction and other parameters.
In an alternative embodiment motion compensation parameter set candidates from base layer blocks are used. The inter prediction is also performed in the base layer (using base layer reference pictures). For applying the error measure the base layer block (b) reconstruction signal can be directly used without upsampling. The selected optimal motion compensation parameter set is applied to enhancement layer reference pictures to form the prediction signal of block (a). When applying motion vectors in a spatial enhancement layer, the motion vectors are scaled according to the resolution change.
Both encoder and decoder can perform the same prediction steps to select the optimal motion compensation parameter set among the available candidates and create identical prediction signals. These parameters are not signaled in the coded video bitstream.
The selection of the prediction method is signaled in the bitstream and can be coded using entropy coding. Inside a hierarchical block subdivision structure this coding method can be selected in every sub-level or alternatively only on subsets of the coding hierarchy.
In an alternative embodiment the encoder can transmit a refinement motion parameter set prediction signal to the decoder. The refinement signal contains differentially coded values of the motion parameters. The refinement signal can be entropy coded.
In an alternative embodiment the decoder creates a list of the best candidates. The index of the used motion parameter set is signaled in the coded video bitstream. The index can be entropy coded. In an example implementation the list can be ordered by increasing error measure.
An example implementation uses the adaptive motion vector prediction (AMVP) candidate list of HEVC to generate motion compensation parameter set candidates.
Another example implementation uses the merge mode candidate list of HEVC to generate motion compensation parameter set candidates.
Method 2: Motion Vector DerivationFor a block (a) of the spatial or quality enhancement layer picture the corresponding block (b) of the base layer picture is determined, that covers the same picture area.
An inter prediction signal for block (a) of the enhancement layer is formed using the following method:
-
- 1. A motion vector predictor is selected.
- 2. A motion estimation on a defined set of search positions is performed on the enhancement layer reference pictures.
- 3. For each search position an error measure is determined and the motion vector with the smallest error is selected.
- 4. A prediction signal for block (a) is formed using the selected motion vector.
In an alternative embodiment the search is performed on the reconstructed base layer signal. For spatial scalability the selected motion vector is scaled according to the spatial resolution change before creating the prediction signal in step 4.
The search positions can be at full or sub-pel resolution. The search can also be performed in multiple steps, e.g. first determining the best full-pel position followed by another set of candidates based on the selected full-pel position. The search can be terminated early, e.g. when the error measure lies below a defined threshold.
Both encoder and decoder can perform the same prediction steps to select the optimal motion vector among the candidates and create identical prediction signals. These vectors are not signaled in the coded video bitstream.
The selection of the prediction method is signaled in the bitstream and can be coded using entropy coding. Inside a hierarchical block subdivision structure this coding method can be selected in every sub-level or alternatively only on subsets of the coding hierarchy.
In an alternative embodiment the encoder can transmit a refinement motion vector prediction signal to the decoder. The refinement signal can be entropy coded.
An example implementation uses the algorithm described in Method 1 to select a motion vector predictor.
Another example implementation uses the adaptive motion vector prediction (AMVP) method of HEVC to select a motion vector predictor from temporally or spatially neighboring blocks of the enhancement layer.
Method 3: Intra Prediction Mode DerivationFor each block (a) in an enhancement layer (n) picture, the corresponding block (b) covering the same area in the reconstructed base layer (n−1) picture is determined.
In a scalable video decoder for each base layer block (b) an intra prediction signal is formed using an intra prediction mode (p) which is inferred by the following algorithm.
-
- 1) An intra prediction signal is created for each available intra prediction mode following the rules for intra prediction of the enhancement layer, but using sample values from the base layer.
- 2) The best prediction mode (pbest) is determined by minimizing an error measure (e.g. sum of absolute differences) between the intra prediction signal and the decoded base layer block (b).
- 3) The prediction (pbest) mode selected in step 2) is used for creating the prediction signal for the enhancement layer block (a) following the intra prediction rules for the enhancement layer
Both encoder and decoder can perform the same steps to select the best prediction mode (pbest) and form a matching prediction signal. The actual intra prediction mode (pbest) is thus not signaled in the coded video bitstream.
The selection of the prediction method is signaled in the bitstream and can be coded using entropy coding. Inside a hierarchical block subdivision structure this coding mode can be selected in every sub-level or alternatively only on subsets of the coding hierarchy.
An alternative embodiment uses samples from the enhancement layer in step 2) to create the intra prediction signal. For a spatial scalable enhancement layer, the base layer can be upsampled using an interpolation filter to apply the error measure.
An alternative embodiment splits the enhancement layer block into multiple blocks of a smaller block size (ai)(e.g. a 16×16 block (a) can be split into 16 4×4 block (ai)). The algorithm described above is applied to each sub-block (ai) and corresponding base layer block (bi)—After prediction of block (ai) a residual coding is applied and the result is used for predicting block (ai+1).
An alternative embodiment uses the surrounding sample values of (b) or (bi) to determine the predicted intra prediction mode (pbest). For instance when a 4×4 block (ai) of a spatial enhancement layer (n) has a corresponding 2×2 base layer block (bi), the surrounding samples of (bi) are used to form a 4×4 block (ci) which is used for determining the predicted intra prediction mode (pbest).
In an alternative embodiment the encoder can transmit a refinement intra prediction direction signal to the decoder. In video codecs, like e.g. HEVC, most intra prediction modes correspond to an angle from which border pixels are used to form the prediction signal. The offset to the optimal mode can be transmitted as difference to the predicted mode (Pbes1) (which is determined as described above). The refinement mode can be entropy coded.
Intra prediction modes are usually coded depending on their probability. In H.264/AVC one most probable mode is determined based on modes used in the (spatial) neighborhood of a block. In HEVC a list most probable modes is created. These most probable modes can be selected using fewer symbols in the bitstream than the whole mode number would necessitate. An alternative embodiment uses the predicted intra prediction mode (pbest) for block (a) (which is determined as described in the algorithm above) as most probably mode or member of the list of most probable modes.
Method 4: Intra Prediction Using Border AreasIn a scalable video decoder to form an intra prediction signal for a block (a) (see
The following alternative variants of selecting these pixels may be used:
-
- a) if the pixel in the surrounding area is not yet coded, the pixel value is not used for predicting the current block
- b) if the pixel in the surrounding area is not yet coded, the pixel value is derived from neighboring pixels that are already coded (e.g. by repetition)
- c) if the pixel in the surrounding area is not yet coded, the pixel value is derived from a pixel in the corresponding area of the decoded base layer picture
To form the intra prediction of the block (a) a neighboring line of pixels (b) (that is derived as described above) is used as a template to fill in each line (aj) of the block (a).
The lines (aj) of block (a) are filled in step-by-step along the x-axis. To achieve the best possible prediction signal, the row of template samples (b) is shifted along the y-axis to form the prediction signal (b′j) for the associated line (aj).
For finding the optimal prediction in each line, the shift offset (oj) is determined by minimizing an error measure between the resulting prediction signal (aj) and the sample values of corresponding line in the base layer.
If (oj) is a non-integer value, an interpolation filter can be used to map the values of (b) to the integer sample positions of (ay) as shown in (b′7).
If spatial scalability is used, an interpolation filter can be used to create a matching number of sample values of the corresponding line of the base layer.
The fill direction (x-axis) can be horizontal (left to right or right to left), vertical (top to bottom or bottom to top), diagonal, or any other angle. The samples used for the template line (b) are the samples in the direct neighborhood of the block along the x-axis. The template line (b) is shifted along the y-axis, which forms a 90° angle to the x-axis.
To find the optimal direction of the x-axis, a full intra prediction signal is created for block (a). The angle with a minimum error measure between the prediction signal and the corresponding base layer block is selected.
The number of possible angles can be restricted.
Both encoder and decoder run the same algorithm to determine the best prediction angles and offsets. No explicit angle or offset information needs to be signaled in the bitstream.
In an alternative embodiment only samples of the base layer picture are used to determine the offsets (oi).
In an alternative embodiment a refinement (e.g. difference value) of the predicted offsets (oi) is signaled in the bitstream. Entropy coding can be used for coding the refinement offset value.
In an alternative embodiment a refinement (e.g. difference value) of the predicted direction is signaled in the bitstream. Entropy coding can be used for coding the refinement direction value.
An alternative embodiment uses a threshold to select if the line (b′j) is used for prediction. If the error measure for the optimal offset (oj) is below the threshold, the line (ci) is used to determine the values of the block line (aj). If the error measure for the optimal offset (oj) is above the threshold, the (upsampled) base layer signal is used to determine the values of the block line (aj)
Method 5: Other Prediction ParametersOther prediction information is inferred similar to methods 1-3, for instance the partitioning of block into sub-blocks:
For a block (a) of the spatial or quality enhancement layer picture the corresponding block (b) of the base layer picture is determined, that covers the same picture area.
A prediction signal for block (a) of the enhancement layer is formed using the following method:
-
- 1) A prediction signal is created for each possible value of the tested parameter.
- 2) The best prediction mode (pbest) is determined by minimizing an error measure (e.g. sum of absolute differences) between the prediction signal and the decoded base layer block (b).
- 3) The prediction (pbest) mode selected in step 2) is used for creating the prediction signal for the enhancement layer block (a)
Both encoder and decoder can perform the same prediction steps to select the optimal prediction mode among the possible candidates and create identical prediction signals. The actual prediction mode is not signaled in the coded video bitstream.
The selection of the prediction method is signaled in the bitstream and can be coded using entropy coding. Inside a hierarchical block subdivision structure this coding method can be selected in every sub-level or alternatively only on subsets of the coding hierarchy.
The following description briefly summarizes some of the above embodiments.
Enhancement Layer Coding with Multiple Methods for Generating an Intra Prediction Signal Using Reconstructed Base Layer Samples
Main aspect: For coding a block in the enhancement layer multiple methods for generating an intra prediction signal using reconstructed base layer samples are provided in addition to methods that generate the prediction signal based on reconstructed enhancement layer samples only.
Sub Aspects:
-
- The multiple methods include the following method: The (upsampled/filtered) reconstructed base layer signal is directly used as enhancement layer prediction signal.
- The multiple methods include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with a spatial intra prediction signal, where the spatial intra prediction is derived based on difference samples for neighbouring blocks. The difference samples represent the difference of the reconstructed enhancement layer signal and the (upsampled/filtered) reconstructed base layer signal (see aspect A).
- The multiple methods include the following method: A conventional spatial intra prediction signal (derived using neighbouring reconstructed enhancement layer samples) is combined with an (upsampled/filtered) base layer residual signal (inverse transform of base layer transform coefficients or difference between base layer reconstruction and base layer prediction)(see aspect B).
- The multiple methods include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with a spatial intra prediction signal, where the spatial intra prediction is derived based on reconstructed enhancement layer samples of neighbouring blocks. The final prediction signal is obtained by weighting the spatial prediction signal and the base layer prediction signal in a way that different frequency components use a different weighting. (see aspect C1) This can be, for example, realized by any of the following:
- Filtering the base layer prediction signal with a low-pass filter and filtering the spatial intra prediction signal with a high-pass filter and adding up the obtained filtered signals. (see aspect C2)
- Transform the base layer prediction signal and the enhancement layer prediction signal and superimpose the obtained transform blocks, where different weighting factors are used for different frequency positions. (see aspect C3) The obtained transform block can then be inverse transformed and used as enhancement layer prediction signal or (see aspect C4) the obtained transform coefficients are added to the scaled transmitted transform coefficient levels and are then inverse transformed in order to obtain the reconstructed block before deblocking and in-loop processing.
- For methods that use the reconstructed base layer signal, the following versions can be used. This can be fixed or it can be signalled at a sequence level, picture level, slice level, largest coding unit level, coding unit level. Or it can be made dependent on other coding parameters.
- Reconstructed base layer samples before deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking but before further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter) or between multiple in-loop processing steps (see aspect D).
- Multiple versions of the methods that use the (upsampled/filtered) base layer signal can be used. The employed upsampled/filtered base layer signal for these versions can differ in the used interpolation filters (including interpolation filters that also filter the integer-sample positions), or the upsampled/filtered base layer signal for the second version can be obtained by filtering the upsampled/filtered base layer signal for the first version. The selection of one of the different version can be signalled at a sequence, picture, slice, largest coding unit, or coding unit level or it can be inferred from the characteristics of the corresponding reconstructed base layer signal or the transmitted coding parameters (see aspect E).
- Different filters can be used for upsampling/filtering the reconstructed base layer signal (see aspect E) and the base layer residual signal (see aspect F).
- For base layer blocks for which the residual signal is zero, it can be replaced with another signal derived from the base layer, e.g., a high-pass filtered version of the reconstructed base layer block (see aspect G).
- For modes that use a spatial intra prediction, non-available neighbouring samples in the enhancement layer (due to a given coding order) can be replaced with the corresponding samples of an upsampled/filtered base layer signal (see aspect H).
- For modes that use a spatial intra prediction, the coding of the intra prediction mode can be modified. The list of most probably modes includes the intra prediction mode of the co-located base layer signal.
- In a particular version, the enhancement layer pictures are decoded in a two-stage process. In a first stage, only the blocks that only use the base layer signal (but do not use neighbouring blocks) or an inter prediction signal for prediction are decoded and reconstructed. In the second stage, the remaining block that use neighbouring samples for prediction are reconstructed. For the blocks that are reconstructed in the second stage, the spatial intra prediction concept can be extended. (see aspect I) Based on the availability of already reconstructed blocks, not only the neighbouring samples to the top and to the left but also neighbouring samples at the bottom and to the right of a current block can be used for spatial intra prediction.
Enhancement Layer Coding with Multiple Methods for Generating an Inter Prediction Signal Using Reconstructed Base Layer Samples
Main aspect: For coding a block in the enhancement layer multiple methods for generating an inter prediction signal using reconstructed base layer samples are provided in addition to methods that generate the prediction signal based on reconstructed enhancement layer samples only.
Sub Aspects:
-
- The multiple methods include the following method: A conventional inter prediction signal (derived by motion-compensated interpolation of already reconstructed enhancement layer pictures) is combined with an (upsampled/filtered) base layer residual signal (inverse transform of base layer transform coefficients or difference between base layer reconstruction and base layer prediction).
- The multiple methods include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with a motion-compensated prediction signal, where the motion-compensated prediction signal is obtained by motion compensating difference pictures. The difference pictures represent the difference of the reconstructed enhancement layer signal and the (upsampled/filtered) reconstructed base layer signal for the reference pictures (see aspect J).
- The multiple methods include the following method: The (upsampled/filtered) reconstructed base layer signal is combined with an inter prediction signal, where the inter prediction is derived by motion-compensated prediction using reconstructed enhancement layer pictures. The final prediction signal is obtained by weighting the inter prediction signal and the base layer prediction signal in a way that different frequency components use a different weighting (see aspect C). This can be, for example, realized by any of the following:
- Filtering the base layer prediction signal with a low-pass filter and filtering the inter prediction signal with a high-pass filter and adding up the obtained filtered signals.
- Transform the base layer prediction signal and the inter prediction signal and superimpose the obtained transform blocks, where different weighting factors are used for different frequency positions. The obtained transform block can then be inverse transformed and used as enhancement layer prediction signal or the obtained transform coefficients are added to the scaled transmitted transform coefficient levels and are then inverse transformed in order to obtain the reconstructed block before deblocking and in-loop processing.
- For methods that use the reconstructed base layer signal, the following versions can be used. This can be fixed or it can be signalled at a sequence level, picture level, slice level, largest coding unit level, coding unit level. Or it can be made dependent on other coding parameters.
- Reconstructed base layer samples before deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking but before further in-loop processing (such as sample adaptive offset filter or adaptive loop filter).
- Reconstructed base layer samples after deblocking and further in-loop processing (such as sample adaptive offset filter or adaptive loop filter) or between multiple in-loop processing steps (see aspect D).
- For base layer blocks for which the residual signal is zero, it can be replaced with another signal derived from the base layer, e.g., a high-pass filtered version of the reconstructed base layer block (see aspect G).
- Multiple versions of the methods that use the (upsampled/filtered) base layer signal can be used. The employed upsampled/filtered base layer signal for these versions can differ in the used interpolation filters (including interpolation filters that also filter the integer-sample positions), or the upsampled/filtered base layer signal for the second version can be obtained by filtering the upsampled/filtered base layer signal for the first version. The selection of one of the different version can be signalled at a sequence, picture, slice, largest coding unit, or coding unit level or it can be inferred from the characteristics of the corresponding reconstructed base layer signal or the transmitted coding parameters (see aspect E).
- Different filters can be used for upsampling/filtering the reconstructed base layer signal (see aspect E) and the base layer residual signal (see aspect F).
- For motion-compensated prediction of difference pictures (difference between enhancement layer reconstruction and upsampled/filtered base layer signal) (see aspect J), different interpolation filters than for motion-compensated prediction of reconstructed pictures are used.
- For motion-compensated prediction of difference pictures (difference between enhancement layer reconstruction and upsampled/filtered base layer signal) (see aspect J), interpolation filters are selected based on characteristic of the corresponding area in the difference picture (or based on coding parameters or based on information transmitted in the bitstream).
Main aspect: Use multiple enhancement layer predictors and at least one predictor derived from the base layer for enhancement layer motion parameter coding.
Sub Aspects:
-
- Addition of (scaled) base layer motion vector to motion vector predictor list (see aspect K)
- Use base layer block that covers co-located sample of centre position of current block (other derivations possible)
- Scale motion vectors according to resolution ratio
- Add motion data of co-located base layer block to merge candidate list (see aspect K)
- Use base layer block that covers co-located sample of centre position of current block (other derivations possible)
- Scale motion vectors according to resolution ratio
- Don't add if merge_flag is equal to 1 in base layer
- Re-ordering of merge candidate list based on base layer merge information (see aspect L)
- If co-located base layer block is merged with a particular candidate, the corresponding enhancement layer candidate is used as first entry in the enhancement layer merge candidate list
- Re-ordering of motion predictor candidate list based on base layer motion predictor information (see aspect L)
- If co-located base layer block uses a particular motion vector predictor, the corresponding enhancement layer motion vector predictor is used as first entry in the enhancement layer motion vector predictor candidate list
- Derivation of merge index (i.e., the candidate with which the current block is merged) based on base layer information in a co-located block (see aspect M). As an example, if the base layer block is merged with a particular neighbouring block and it is signalled inside the bitstream that the enhancement layer block is also merged, no merge index is transmitted, but instead the enhancement layer block is merged with the same neighbour (but in the enhancement layer) as the co-located base layer block.
- Addition of (scaled) base layer motion vector to motion vector predictor list (see aspect K)
Main aspect: Inference of enhancement layer partitioning and motion parameters based on base layer partitioning and motion parameters (probably necessitated to combine this aspect with any of the sub aspects).
Sub Aspects:
-
- Derive motion parameters for N×M sub-blocks of an enhancement layer based on the co-located base layer motion data; summarize block with identical derived parameters (or parameters with small differences) to larger blocks; determine prediction and coding units (see aspect T).
- Motion parameters may include: number of motion hypotheses, reference indices, motion vectors, motion vector predictor identifiers, merge identifiers.
- Signal one of multiple methods for generating the enhancement layer prediction signal; such methods can include the following:
- Motion compensation using the derived motion parameters and the reconstructed enhancement layer reference pictures.
- Combination of (a) (upsampled/filtered) base layer reconstruction for current picture and (b) motion compensated signal using the derived motion parameters and enhancement layer reference picture that are generated by subtracting the (upsampled/filtered) base layer reconstruction from the reconstructed enhancement layer picture.
- Combination of (a) (upsampled/filtered) base layer residual (difference between reconstructed signal and prediction or inverse transform of coded transform coefficient values) for current picture and (b) motion compensated signal using the derived motion parameters and the reconstructed enhancement layer reference pictures.
- If the co-located block in the base layer is intra-coded, the corresponding enhancement layer M×N block (or CU) is also intra coded, where the intra prediction signal is derived by using base layer information (see aspect U), for example:
- An (upsampled/filtered) version of the corresponding base layer reconstruction is used as intra prediction signal (see aspect U).
- The intra prediction mode is derived based on the intra prediction mode used in the base layer and this intra prediction mode is used for spatial intra prediction in the enhancement layer.
- If the co-located base layer block for an M×N enhancement layer block (subblock) is merged with a previously coded base layer block (or has the same motion parameters), the M×N enhancement layer (sub-)block is also merged with the enhancement layer block that corresponds to the base layer block that is used for merging in the base layer (i.e., the motion parameters are copied from the corresponding enhancement layer block)(see aspect M).
Main aspect: Transform coefficient coding using different scan patterns. For enhancement layers, context modelling based on coding mode and/or base layer data, and different initializations for context models.
Sub Aspects:
-
- Introduce one or more additional scan patterns, for example a horizontal and vertical scan pattern. Redefine the subblocks for the additional scan pattern. Instead of 4×4 subblocks, for example 16×1 or 1×16 subblocks can be used, or 8×2 and 8×2 subblocks can be used. The additional scan pattern can be introduced only for blocks greater than or equal to a particular size, e.g. 8×8 or 16×16 (see aspect V).
- The selected scan pattern is signalled inside the bitstreams (if the coded block flag is equal to 1)(see aspect N). For signalling the corresponding syntax element, fixed context can be used. Or the context derivation for the corresponding syntax elements can depend on any of the following:
- Gradient of the co-located reconstructed base layer signal or the reconstructed base layer residual. Or detected edges in the base layer signal.
- Transform coefficient distribution in the co-located base layer blocks.
- The selected scan can be directly derived from the base layer signal (without transmitted any additional data) based on the properties of the co-located base layer signal (see aspect N):
- Gradient of the co-located reconstructed base layer signal or the reconstructed base layer residual. Or detected edges in the base layer signal.
- Transform coefficient distribution in the co-located base layer blocks.
- The different scans can be realized in a way that the transform coefficients ae re-ordered after quantization at the encoder side and the conventional coding is used. At the decoder side, the transform coefficients are conventionally decoded and re-ordered before the scaling and inverse transform (or after the scaling and before the inverse transform).
- For coding the significance flags (sub-group flags and/or significance flags for single transform coefficients), the following modifications may be used in an enhancement layer
- Separate context models are used for all or a subset of the coding modes that use base layer information. It is also possible to use different context models for different modes with base layer information.
- The context modelling can depend on the data in the co-located base layer block (e.g., number of significant transform coefficients for particular frequency positions) (see aspect 0).
- A generalized template can be used that evaluated both, the number of significant already coded transform coefficient levels in a spatial neighbourhood of the coefficient to be coded and the number of significant transform coefficients in the co-located base layer signal of similar frequency positions (see aspect 0).
- For coding the last significant scanning position, the following modifications may be used in an enhancement layer:
- Separate context models are used for all or a subset of the coding modes that use base layer information. Itis also possible to use different context models for different modes with base layer information (see aspect P).
- The context modelling can depend on the data in the co-located base layer block (e.g., transform coefficient distribution in base layer, gradient information of base layer, last scanning position in co-located base layer blocks).
- The last scanning position can be coded as difference to the last base layer scanning position (see aspect S).
- Usage of different context initialization tables for base and enhancement layer.
Main aspect: Use base layer data for deriving enhancement layer coding parameters.
Sub Aspects:
-
- Derive merge candidate based on the (potentially upsampled) base layer reconstruction. In the enhancement layer, only the usage of merge is signalled, but the actually candidate that is used for merging the current block is derived based on the reconstructed base layer signal. Therefore, for all merge candidates, an error measure between the (potentially upsampled) base layer signal for the current enhancement layer block and the corresponding prediction signals (derived using the motion parameters for the merge candidates) is evaluated for all merge candidates (or a subset thereof) and the merge candidate that is associated with the smallest error measure is selected. The calculation of the error measure can also be done in the base layer using the reconstructed base layer signal and the base layer reference pictures (see aspect Q).
- Derive motion vectors based on the (potentially upsampled) base layer reconstruction. Motion vector differences are not coded, but inferred based on the reconstructed base layer. Determine a motion vector predictor for the current block and evaluate a defined set of search positions around the motion vector predictor. For each search position, determine an error measure between the (potentially upsampled) base layer signal for the current enhancement layer block and the displaced reference frame (the displacement is given by the search position). Chose the search position/motion vector that yields the smallest error measure. The search can be split in several stages. For example, a full-pel search is carried out first, followed by a half-pel search around the best ful-pel vector, followed by a quarter-pel search around the best full/half-pel vector. The search can also be carried out in the base layer using the reconstructed base layer signal and the base layer reference pictures, the found motion vectors are then scaled according to the resolution change between base and enhancement layer (see aspect Q).
- Derive intra prediction modes based on the (potentially upsampled) base layer reconstruction. Intra prediction modes are not coded, but inferred based on the reconstructed base layer. For each possible intra prediction mode (or a subset thereof), determine an error measure between the (potentially upsampled) base layer signal for the current enhancement layer block and the intra prediction signal (using the tested prediction mode). Chose the prediction mode that yields the smallest error measure. The calculation of the error measure can also be done in the base layer using the reconstructed base layer signal and the intra prediction signal in the base layer. Furthermore, an intra block can be implicitly decomposed into 4×4 blocks (or other block sizes) and for each 4×4 block a separate intra prediction mode can be determined (see aspect Q).
- The intra prediction signal can be determined by a row-wise or column-wise matching of the border samples with the reconstructed base layer signal. For deriving a shift between the neighbouring samples and a current line/row, an error measure is calculated between the shifted line/row of neighbouring samples and the reconstructed base layer signal, and the shift that yields the smallest error measure is chosen. As neighbouring samples, the (upsampled) base layer samples or the enhancement layer samples can be used. The error measure can also be directly calculated in the base layer (see aspect W).
- Using backward-adaptive method for the derivation of other coding parameters such as block partitioning, etc.
A further brief summary of the above embodiments is presented below. In particular, above embodiments described,
A1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a, 200b, 200c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- subject (220) the reconstructed base layer signal (200a, 200b, 200c) to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- compute (260) a difference signal between an already reconstructed portion (400a or 400b) of an enhancement layer signal and the inter-layer prediction signal (380);
- spatially predict (260) the difference signal at a first portion (440, cp.
FIG. 46 ) collocated to a portion of the enhancement layer signal (360), currently to be reconstructed, from a second portion (460) of the difference signal, spatially neighboring the first portion and belonging to the already reconstructed portion of the enhancement layer signal (360) to obtain a spatial intra prediction signal; - combine (260) the inter-layer prediction signal (380) and the spatial intra prediction signal to obtain an enhancement layer prediction signal (420); and
- predictively reconstructing (320, 580, 340, 300, 280) the enhancement layer signal (360) using the enhancement layer prediction signal (420).
In accordance with the aspect A1, the base layer signal may be reconstructed by the base layer decoding stage 80 from the coded data stream 6 or the substream 6a, respectively, in the block-based predictive manner described above with, for example, transform decoding as far as the base layer residual signal 640/480 is concerned, but other reconstruction alternatives are also feasible.
As far as the reconstruction of the enhancement layer signal 360 by the enhancement layer decoding stage 60 is concerned, the resolution or quality refinement to which the reconstructed base layer signal 200a, 200b, or 200c is subject may, for example, involve up-sampling in the case of a resolution refinement, or copying in case of quality refinement, or tone-mapping from n bits to m bits with m>n in case of bit depth refinement.
The computation of the difference signal may be done pixel-wise, i.e., co-located pixels of the enhancement layer signal on the one hand and the prediction signal 380 on the other hand are subtracted from each other and this is done per pixel position.
The spatial prediction of the difference signal may be done in any way, such as by transmitting in the coded data stream 6 or within substream 6b an intra-prediction parameter, such as an intra prediction direction, and copying/interpolating already reconstructed pixels bordering the portion of the enhancement layer signal 360, currently to be reconstructed, along this intra prediction direction into the current portion of the enhancement layer signal. The combination may involve a summation, a weighted sum or even more sophisticated combinations such as combinations which differently weight the contributions in the frequency domain.
The predictive reconstruction of the enhancement layer signal 360 using the enhancement layer prediction signal 420 may, as shown in the figure, involve the entropy decoding and inverse transformation of an enhancement layer residual signal 540 and a combination 340 of the latter with the enhancement layer prediction signal 420.
B1) Scalable video decoder configured to
-
- decode (100) a base layer residual signal (480) from a coded data stream (6),
- reconstruct (60) the enhancement layer signal (360) comprising
- subjecting (220) the reconstructed base layer residual signal (480) to a resolution or quality refinement to obtain an inter-layer residual prediction signal (380),
- spatially predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) to obtain an enhancement layer internal prediction signal;
- combining (260) the inter-layer residual prediction signal and the enhancement layer internal prediction signal to obtain an enhancement layer prediction signal (420); and
- predictively reconstructing (340) the enhancement layer signal (360) using the enhancement layer prediction signal (420).
The decoding of the base layer residual signal from the coded data stream may, as shown in the figure, be performed by using entropy decoding and inverse transformation. Moreover, the scalable video decoder may also, optionally, perform the reconstruction of the base layer signal itself, namely by predictive decoding by deriving a base layer prediction signal 660 and combining same with the base layer residual signal 480. As just mentioned, this is merely optional.
As far as the reconstruction of the enhancement layer signal is concerned, the resolution or quality refinement may be performed as denoted above with respect to A).
As far as the spatial prediction of the portion of the enhancement layer signal is concerned, this spatial prediction may also be performed as exemplarily outlined in A) with respect to the difference signal. A similar note is valid as far as the combination and the predictive reconstruction is concerned.
However, it should be mentioned that the base layer residual signal 480 in aspect B is not restricted to equal the explicitly signaled version of the base layer residual signal 480. Rather, it may be possible that the scalable video decoder subtracts any reconstructed base layer signal version 200 with the base layer prediction signal 660, thereby obtaining abase layer residual signal 480 which may deviate from the explicitly signaled one by deviations stemming from filter functions such as of filters 120 or 140. The latter statement is also valid for other aspects where the base layer residual signal is involved in inter-layer prediction.
C1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a. 200b; 200c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- subjecting (220) the reconstructed base layer signal (200) to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- spatially or temporally predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion (400a,b in case of“spatially”; 400a,b,c in case of “temporally”) of the enhancement layer signal (360) to obtain an enhancement layer internal prediction signal;
- forming (260), at the portion currently to be reconstructed, a weighted average of the inter-layer prediction signal and the enhancement layer internal prediction signal (380) to obtain an enhancement layer prediction signal (420) such that the weights at which the inter-layer prediction signal and the enhancement layer internal prediction signal (380) contribute to the enhancement layer prediction signal (420) vary over different spatial frequency components; and
- predictively reconstructing (320, 340) the enhancement layer signal (360) using the enhancement layer prediction signal (420).
C2) Wherein the formation (260) of the weighted average comprises, at the portion currently to be reconstructed, filtering (260) the inter-layer prediction signal (380) with a low-pass filter and filtering (260) the enhancement layer internal prediction signal with a high-pass filter to obtain filtered signals, and adding-up the obtained filtered signals.
C3) Wherein the formation (260) of the weighted average comprises, at the portion currently to be reconstructed, transforming (260) the inter-layer prediction signal and the enhancement layer internal prediction signal so as to obtain transform coefficients; and superimposing (260) the obtained transform coefficients with using different weighting factors for different spatial frequency components to obtain superimposed transform coefficients; and inverse transforming the superimposed transform coefficients to obtain the enhancement layer prediction signal.
C4) Wherein the predictive reconstruction (320, 340) of the enhancement layer signal using the enhancement layer prediction signal (420) comprises extracting (320) transform coefficient levels for the enhancement layer signal from a coded data stream (6), performing (340) a sum of the transform coefficient levels and the superimposed transform coefficients so as to obtain a transformed version of the enhancement layer signal and subjecting the transformed version of the enhancement layer signal to an inverse transform so as to obtain the enhancement layer signal (360) (i.e. the inverse transform T−1 in the figure would be placed downstream the adder 340, at least for that coding mode).
As far as the reconstruction of the base layer signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspects A) and B).
The same applies to the resolution or quality refinement mentioned in C, as well as the spatial prediction.
The temporal prediction mentioned in C may involve the prediction provider 160 deriving motion prediction parameters from the coded data stream 6 and substream 6a, respectively. The motion parameters may comprise: a motion vector, a reference frame index, or they may comprise a combination of a motion subdivision information and a motion vector per sub-block of the currently reconstructed portion.
As described before, the formation of the weighted average may end up in the spatial domain or the transform domain and accordingly, the adding at adder 340 may be performed in the spatial or transform domain. In the latter case, the inverse transformer 580 would apply the inverse transform onto the weighted average.
D1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a,b, 200c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (380) comprising
- subjecting (220) the reconstructed base layer signal to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- predictively reconstructing (320, 340) the enhancement layer signal (360) using the inter-layer prediction signal (380),
- wherein the reconstruction (60) of the enhancement layer signal is performed such that the inter-layer prediction signal (380) evolves, controlled via side-information in the coded bit-stream 360), from a different one of none (200a), one or all (200b,c) of a deblocking and an in-loop filtering (140) for different portions of a video scalably represented by the base layer signal and enhancement layer signal, respectively.
As far as the reconstruction of the base layer signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspects A) and B). The same applies to the resolution or quality refinement.
The predictive reconstruction mentioned in D may involve the prediction provider 160, as described above, and may involve
-
- spatially or temporally predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (380) to obtain an enhancement layer internal prediction signal;
- combining (260) the inter-layer prediction signal (380) and the enhancement layer internal prediction signal to obtain the enhancement layer prediction signal (420).
The fact that the inter-layer prediction signal (380) evolves, controlled via side-information in the coded bit-stream (360), from a different one of none (200a), one or all (200b,c) of a deblocking and an in-loop filtering (140) for different portions of the video means the following:
Naturally, the base layer substream 6a itself may (optionally) signal the usage of different measures to result in the final base layer signal 600 such as bypassing all filters 120, 140, the use of merely deblocking or the use of merely in-loop filtering or the use of both deblocking and in-loop filtering. Even the filter transfer functions maybe signaled/varied by side information in 6a. The granularity defining the different portions at which these variations are done may be defined by the afore-mentioned coding units, prediction blocks or any other granularity. The scalable video decoder (the coding stage 80) thus applies these variations if merely the base layer signal is to be reconstructed. However, independent therefrom, substream 6b comprises side-information, which signals a new (i.e. independent from the just-mentioned side-information in base layer signal 6a) variation of which combination of filtering is used to obtain the base layer signal which, then, is used in the predictive reconstruction of the enhancement signal: bypassing all filters 120, 140, the use of merely deblocking or the use of merely in-loop filtering or the use of both deblocking and in-loop filtering. Even the filter transfer functions may be signaled/varied by side information in 6b. The granularity defining the different portions at which these variations are done may be defined by the afore-mentioned coding units, prediction blocks or any other granularity and may differ from the granularity at which this signaling is used in the base layer signal 6a.
E1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a,b,c) from a coded data stream (6);
- reconstruct (60) an enhancement layer signal (360) comprising
- subjecting (220) the reconstructed base layer signal to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- predictively reconstructing (320, 340) the enhancement layer signal (60) using the inter-layer prediction signal (380),
- wherein the reconstruction (60) of the enhancement layer signal (360) is performed such that the inter-layer prediction signal evolves, controlled via side-information in the coded bit-stream (6) or signal-dependent, from different filter transfer functions for an upsampling interpolation filter (220) for different portions of a video scalably represented by the base layer signal and enhancement layer signal, respectively.
As far as the reconstruction of the base layer signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspects A) and B). The same applies to the resolution or quality refinement.
The predictive reconstruction mentioned may involve the prediction provider 160, as described above, and may involve
-
- spatially or temporally predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) to obtain an enhancement layer internal prediction signal;
- combining (260) the inter-layer prediction signal (380) and the enhancement layer internal prediction signal to obtain an enhancement layer prediction signal (420).
The fact that the inter-layer prediction signal evolves, controlled via side-information in the coded bit-stream (6) or signal-dependent, from different filter transfer functions for an upsampling interpolation filter (220) for different portions of the video means the following:
Naturally, the base layer substream 6a itself may (optionally) signal the usage of different measures to result in the final base layer signal 600 such as bypassing all filters 120, 140, the use of merely deblocking or the use of merely in-loop filtering or the use of both deblocking and in-loop filtering. Even the filter transfer functions maybe signaled/varied by side information in 6a. The granularity defining the different portions at which these variations are done may be defined by the aforementioned coding units, prediction blocks or any other granularity. The scalable video decoder (the coding stage 80) thus applies these variations if merely the base layer signal is to be reconstructed. However, independent therefrom, substream 6b may comprise side-information, which signal additionally (i.e. independent from the just-mentioned side-information in base layer signal 6a) a variation of the filter transfer function used in refiner 220 to obtain refined signal 380. The granularity defining the different portions at which these variations are done may be defined by the afore-mentioned coding units, prediction blocks or any other granularity and may differ from the mentioned granularity of the base layer signal 6a.
As described above, the variation to be used may be inferred signal-dependent, with or without using additional side information, from the base-layer signal, base layer residual signal or the coding parameters in substream 6a.
F1) Scalable video decoder configured to
-
- decode (100) a base layer residual signal (480) from a coded data stream,
- reconstruct (60) an enhancement layer signal (360) by subjecting (220) the reconstructed base layer residual signal (480) to a resolution or quality refinement to obtain an inter-layer residual prediction signal (380) and predictively reconstructing (320,340, and, optionally, 260) the enhancement signal (360) using the inter-layer residual prediction signal (380),
- wherein the reconstruction (60) of the enhancement layer signal (360) is performed such that the inter-layer residual prediction signal evolves, controlled via side-information in the coded bit-stream or signal-dependent, from different filter transfer functions for different portions of a video scalably represented by the base layer signal and enhancement layer signal, respectively.
As far as the reconstruction of the base layer residual signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspect B). The same applies to the resolution or quality refinement.
The predictive reconstruction mentioned may involve the prediction provider 160, as described above, and may involve
-
- spatially or temporally predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) to obtain an enhancement layer internal prediction signal;
- decoding (320) an enhancement residual signal from the coded data stream;
- combining (involving also 340 and 260) the enhancement layer internal prediction signal, the inter-layer residual prediction signal (380) and the enhancement layer residual signal to obtain the enhancement layer signal (360).
The fact that the inter-layer residual prediction signal evolves, controlled via side-information in the coded bit-stream, from different filter transfer functions for different portions of the video means the following:
Naturally, the base layer substream 6a itself may (optionally) signal the usage of different measures to result in the final base layer signal 600 such as bypassing all filters 120, 140. See above descriptions of D) and E). However, independent therefrom, substream 6b may comprise side-information, which signal additionally (i.e. independent from the just-mentioned side-information in base layer signal 6a) a variation of the filter transfer function used in refiner 220 to obtain refined residual signal 380. The granularity defining the different portions at which these variations are done may be defined by the afore-mentioned coding units, prediction blocks or any other granularity and may differ from the mentioned granularity of the base layer signal 6a.
As described above, the variation to be used may be inferred signal-dependent, with or without using additional side information, from the base-layer signal, base layer residual signal or the coding parameters in substream 6a.
G1) Scalable video decoder configured to
-
- decode (100) abase layer residual signal (480) of a base layer signal (200) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) by subjecting (220) the reconstructed base layer residual signal (480) to a resolution or quality refinement to obtain an inter-layer residual prediction signal (380) and predictively reconstructing (320,340, and, optionally, 260) the enhancement layer signal (360) using the inter-layer residual prediction signal (480),
- wherein the reconstruction (60) of the enhancement layer signal (360) comprises identifying (260) blocks of the base layer residual signal (480) being all zero and replacing the identified blocks with a replacement signal derived from a portion of the base layer signal (200).
As far as the decoding of the base layer residual signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspect B). The same applies to the resolution or quality refinement.
The predictive reconstruction mentioned may involve the prediction provider 160, as described above, and may involve
-
- spatially or temporally predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) to obtain an enhancement layer internal prediction signal;
- decoding (320) an enhancement residual signal from the coded data stream;
- combining (involving also 340 and 260) the enhancement layer internal prediction signal, the inter-layer residual prediction signal (380) and the enhancement layer residual signal to obtain the enhancement layer signal (360).
Advantageously, the identification involves the provider 260 checking a syntax element in the base layer stream (6a) indicating, at a granularity of transform blocks, for example, which are, for example, a further subdivision of the afore-mentioned coding units, as to whether the respective transform block is all zero or not.
The portion of the base layer signal used for replacement may, as described above, for example, be a high-pass filtered version of the base layer signal 200a,b,c.
H1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a,b,c) from a coded data stream (6);
- reconstruct (60) an enhancement layer signal comprising
- subjecting (220) the reconstructed base layer signal (200) to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- spatially predicting (260) a portion of an enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) to obtain an enhancement layer prediction signal (420);
- predictively reconstructing (340) the enhancement layer signal (360) using the enhancement layer prediction signal (420),
- wherein the spatially predicting (260) comprises replacing a non-available portion within an predetermined neighborhood of the portion currently to be reconstructed, with a co-located portion, co-located to the non-available portion, of the inter-layer prediction signal (380) und performing the spatial prediction also dependent on the co-located portion of the inter-layer prediction signal (380).
As far as the reconstruction of the base layer signal is concerned, reference is made to the above descriptions, such as with respect to the figure in general and with respect to aspects A) and B). The same applies to the resolution or quality refinement, as well as the spatial prediction and the predictive reconstruction which were also described above.
The non-availability may result from the following facts: imagine that the predetermined neighborhood encompasses neighboring samples above the top edge of the portion/block currently to be reconstructed as well as samples extending further to the right thereof, and samples to the left of the left edge of the current block/portion currently to be reconstructed as well as samples extending further downwards thereof. Further, imagine that the spatial prediction is applied to blocks/portions being the leafs of a multi-tree based subdivision and that these blocks are traversed in the predictive reconstruction process in a depth-first-zig-zag-traversal order. Then, some blocks have all neighboring samples in question available, i.e. they are already reconstructed. Some blocks, however, lack some of these samples, i.e. they are not completely reconstructed. The missing samples are then replaced as described. Another fact potentially necessitating replacement is the position of slice borders in internal to any frame. The replacement itself is done by copying the co-located samples of the inter-layer prediction signal (380) to the missing samples. Then, the spatial prediction is done using the complete (completed) predetermined neighborhood, including samples copied from the inter-layer prediction signal (380) and samples from the reconstructed enhancement layer signal.
I1) Scalable video decoder configured to
-
- reconstruct (60) an enhancement layer signal (360) by using, for different blocks of a frame, different ones of an inter-layer prediction mode, a temporal inter prediction mode and a spatial intra-prediction mode,
- wherein the scalable video decoder is configured to, in reconstructing (60) an enhancement layer signal (360), reconstruct (60) the blocks by traversing the blocks in a first scan and then in a second scan, and, in the first scan, skip blocks to be reconstructed using the spatial intra-prediction mode and reconstruct blocks to be reconstructed using one of the inter-layer prediction mode and the inter prediction mode, and, in the second scan, reconstruct the blocks to be reconstructed using the spatial intra-prediction mode.
As to “spatial intra prediction mode” and “temporal inter-prediction mode” reference is made to the above discussion for possible implementations. As to “inter-layer prediction mode” any of the examples for such modes may be used as described so far or below. Due to the increased chance of having the complete outline of the spatial intra-prediction mode blocks of the enhancement layer signal surrounded by already reconstructed samples, the template of neighboring samples from which the inner of the currently predicted spatial intra-prediction mode block is to be filled, may be increased to completely surround the outline.
J1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a,b,c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- subjecting (220) the reconstructed base layer signal (200a,b,c) to a resolution or quality refinement to obtain an inter-layer prediction signal (380) for a reference frame and a current frame,
- forming (260) a difference signal between the inter-layer prediction signal (380) for a reference frame and the already reconstructed enhancement layer signal (360) for the reference frame;
- subjecting (260) the difference signal to motion compensated prediction so as to obtain a difference signal prediction for the current frame;
- combining (260) the inter-layer prediction signal (380) for the current frame and the difference signal prediction for the current frame so as to obtain an enhancement layer prediction signal (420); and
- predictively reconstructing (320, 340, 300, 280) the enhancement layer signal (360) using the enhancement layer prediction signal (420).
J2) Wherein in forming the difference signal for the reference frame, the used reconstructed enhancement layer signal is the enhancement layer reconstruction before deblocking, after deblocking but before optional loop filtering, or after deblocking and optional loop filtering.
J3) Wherein the selection of the enhancement layer signal that is used for forming the difference signal of the reference frame is signaled inside the bitstream at a sequence, picture, or block level.
The motion compensated prediction may involve the prediction provider 260 deriving motion prediction parameters from the coded data stream 6 and substream 6b, respectively, for a currently reconstructed block of the enhancement signal, the derivation involving the decoder 320. The motion vector(s) is/are applied to the location of the portion of the enhancement signal currently to be reconstructed, and the respectively displaced portion(s) is/are copied (with or without fractional-sample interpolation) from the difference signal. The thus copied portion is, in the combination, for example, pixel-wise added with the portion of the inter-layer prediction signal (380), co-located to the currently to be reconstructed portion.
K1) Scalable video decoder configured to
-
- decode base layer motion parameters (520) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (260) comprising
- gathering (260), for a block of a frame of the enhancement layer signal (360), a set of motion parameter candidates from neighboring already reconstructed blocks of the frame;
- gathering (260) base layer motion parameters of a block of the base layer signal, collocated to the block of the frame of the enhancement layer signal (360), from the base layer motion parameters (520);
- adding (260) the base layer motion parameters or a scaled version of the base layer motion parameters to the set of motion parameter candidates so as to obtain an extended motion parameter candidate set of motion parameter candidates,
- selecting (260) at least one of the motion parameter candidates of the extended motion parameter candidate set,
- predicting (260), by motion compensated prediction, the enhancement layer signal using the selected one of the motion parameter candidates of the extended motion parameter candidate set.
K2) Wherein the scalable video decoder is configured to scale the base layer motion parameters according to a spatial resolution ratio between the base layer signal and the enhancement layer signal to obtain the scaled version of the base layer motion parameters.
K3) Wherein the scalable video decoder is configured to check as to whether the base layer motion parameters are coded in the coded data stream using merging or not, and, if the base layer motion parameters are coded in the coded data stream using merging, suppress the adding.
The motion parameters mentioned in this aspect may pertain to the motion vectors only (motion vector prediction), or to the complete set of motion parameters including the number of motion hypotheses per block, reference indices, partitioning information (merging).
Accordingly, the “scaled version” may stem from a scaling of the motion parameters used in the base layer signal in accordance with the spatial resolution ratio between base and enhancement layer signal in case of spatial scalability.
The decoding of the base layer motion parameters 520 from the coded data stream, may involve the motion vector prediction or merging also.
The incorporation of the motion parameters used at a co-located portion of the base layer signal into the set of merging/motion vector candidates enables a very effective indexing among the intra-layer candidates and the inter-layer candidate.
The selection may involve explicit signaling of an index into the extended set/list of motion parameter candidates in the enhancement layer signal such as for a prediction block, a coding unit or the like. Alternatively, the selection index may be inferred from other information of the enhancement layer signal 6b or inter-layer information.
L1) Scalable video decoder configured to
-
- decode (100) base layer motion parameters (520) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- ordering (240) a motion parameter candidate list for the enhancement layer signal depending on the base layer motion parameters;
- selecting (240), controlled via an explicitly signaled index syntax element in the coded data stream (6), enhancement layer motion parameters from the ordered motion parameter candidate list for the enhancement layer signal, and
- predicting (260), by motion compensated prediction, the enhancement layer signal using the determined motion parameter.
As to the motion parameters mentioned in this aspect, the same as said above with respect to aspect K applies.
The decoding of the base layer motion parameters 520 from the coded data stream, same may (optionally) involve the motion vector prediction or merging also.
The ordering may be done in accordance with a measure which measures the difference between the respective enhancement layer motion parameter candidates and the base layer motion parameters of the base layer signal, relating to a block of the base layer signal co-located to a current block of the enhancement layer signal. That is, for a current block of the enhancement layer signal, the list of enhancement layer motion parameter candidates may be determined first. Then, the ordering is performed is just-described. Thereinafter, the selection is performed by explicit signaling.
The ordering may also be done in accordance with a measure which measures the difference between the base layer motion parameters of the base layer signal, relating to a block of the base layer signal co-located to a current block of the enhancement layer signal, and the base layer motion parameters of spatially and/or temporally neighboring blocks in the base layer. The determined ordering in the base layer is then transferred to the enhancement layer, so that the enhancement layer motion parameter candidates are ordered in a way that the ordering is the same as the determined ordering for the corresponding base layer candidates, where a base layer motion parameter candidate is said to correspond to an enhancement layer motion parameter candidate when the associated base layer block is spatially/temporally co-located with the enhancement layer block that is associated with the considered enhancement layer motion parameters candidate. Based on the ordering, the selection is performed by explicit signaling.
M1) Scalable video decoder configured to
-
- decode (100) base layer motion parameters (520) from a coded data stream (6) using an index into a motion parameter candidate list for a base layer signal (200),
- reconstruct (60) an enhancement layer signal (360) comprising
- determining (240) an index into a motion parameter candidate list for the enhancement layer signal depending on the index into the motion parameter candidate list for the base layer signal;
- determining (240) an enhancement layer motion parameter using the index into the motion parameter candidate list for the enhancement layer signal, and
- predicting (260), by motion compensated prediction, the enhancement layer signal using the determined motion parameter.
As to the motion parameters mentioned in this aspect, the same as said above with respect to aspect K applies.
The decoding of the base layer motion parameters 520 may involve, for blocks of the base layer signal:
Inspecting a flag signaling as to whether the motion parameter for the current block is signaled in the base layer substream 6a by way of merging or without merging (predictively or independently coded),
If coded using merging or coded predictively,
-
- Determining a list of base layer motion parameters; for example, the motion parameters having been used/selected for neighboring blocks of the base layer signal are used for the determination;
- an index into the respective list is obtained from the base layer signal 6a and used to select one of the base layer motion parameters in the list;
- the index for the enhancement layer is determined in a way that the enhancement layer block co-located with the base layer block associated with the indexed base layer candidate is selected.
For a current block of the enhancement layer signal, a list of enhancement layer motion parameters is determined; for example, the motion parameters having been used/selected for neighboring blocks of the enhancement layer signal are used for the determination.
The selection of the motion parameter index in the enhancement layer may be done in accordance with a measure which measures the difference between the respective enhancement layer motion parameters and the base layer motion parameter of (i.e. having been used or selected in) the base layer signal, relating to a block of the base layer signal co-located to the current block of the enhancement layer signal.
The selection of the motion parameter index in the enhancement layer may also be done in accordance with a measure which measures the difference between the base layer motion parameters of the base layer signal, relating to a block of the base layer signal co-located to a current block of the enhancement layer signal, and the base layer motion parameters of spatially and/or temporally neighboring blocks in the base layer. The selection for the base layer is then transferred to the enhancement layer, so that the enhancement layer motion parameter candidate that corresponds to the selected base layer candidate is chosen, where a base layer motion parameter candidate is said to correspond to an enhancement layer motion parameter candidate when the associated base layer block is spatially/temporally co-located with the enhancement layer block that is associated with the considered enhancement layer motion parameters candidate.
N1) Scalable video decoder configured to
-
- decode (100) a base layer residual signal (480) of a base layer signal from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- determining (240) a gradient of, or information on a spectral decomposition of, the base layer residual signal (480) or the base layer signal to obtain a scan predictor (520);
- decode (320)transform coefficients of an enhancement layer residual signal (540) from the coded data stream (6) using a scan pattern which depends on the scan predictor (520).
That is, the decoding of the transform coefficients may involve a selection of the scan pattern out of a set of possible scan patterns all of which completely cover the transform block. Advantageously, the selection is made such that the scan pattern selected traverses the significant spectral components of the base layer residual signal earlier than the other scan patterns of the set of possible scan patterns.
R1) Video decoder configured to decode (320) transform coefficients of transform block of a residual signal (540 or output of 100) from a coded data stream (6) by selecting a scan pattern out of a set of possible scan patterns all of which completely cover the transform block, the selection being dependent on an explicit signaling in the coded data stream.
R2) Wherein decoding (320) uses a context model for syntax elements related to transform coefficients which depends on (via 520 and 240) the selection of the scan pattern.
R2a) Wherein the syntax elements related to transform coefficients include a syntax element that indicates whether a transform coefficient is equal or unequal to zero for a particular scan position.
R2b) Wherein the syntax elements related to transform coefficients include a syntax element that indicates the position of the last non-zero transform coefficient in the given scan order.
R2c) Wherein the syntax elements related to transform coefficients include a syntax element that indicates whether a subblock of the transform block contains transform coefficients unequal to zero.
R3) Wherein the explicit signaling involves entropy decoding (320) an index into the set of possible scan patterns using a context model which depends on a gradient of, or information on a spectral decomposition of, the base layer residual signal (480) or the base layer signal.
R4) Wherein for decoding (320) the transform coefficient levels the transform block is subdivided into subblocks, a syntax element is transmitted that signals whether a subblock contains non-zero-transform coefficients, and the size or form of the subblocks or the partitioning of the transform block into subblocks is dependent on the selected scan pattern.
S1) Scalable video decoder configured to
-
- decode (100) transform coefficients of transform block of a base layer residual signal from a coded data stream (6) including decoding a first syntax element from the coded data stream which indicates a position of a last significant transform coefficient of the transform block of the base layer residual signal; and
- decode (100) transform coefficients of transform block of an enhancement layer residual signal from the coded data stream (6) including a second decoding syntax element from the coded data stream and computing a position of a last significant transform coefficient of the transform block of the enhancement layer residual signal based on the first syntax element and the second syntax element.
01) Scalable video decoder configured to
-
- decode (100) a base layer residual signal (480) of a base layer signal from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- decoding (320) a syntax element relating to a transform coefficient block of the enhancement layer residual signal (540) from the coded data stream (6) using a context model or a predictor which depends on (via 520 and 240) the base layer residual signal (480) or the base layer signal.
For example, a template is used to determine a context for coding a certain transform coefficient at a currently visited transform coefficient position, and the template also involves positions in one or more transform blocks in the base layer residual signal which are correspond to the certain transform coefficient position in terms of location (at granularity of the transform block sizes in base and enhancement layer) and spectral frequency.
Or, the gradient of, or information on a spectral decomposition of, the base layer residual signal (480) or the base layer signal is used to determine the context model.
P1) Scalable video decoder configured to
-
- decode (100) a base layer residual signal (480) of a base layer signal from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- predict (260) portions of an enhancement layer signal (360) from already reconstructed portions of the enhancement layer signal (360), the base layer residual signal and the base layer signal in accordance with prediction modes which differ for the portions of the enhancement layer signal (360);
- decoding (320) a syntax element relating to a transform coefficient block of an enhancement layer residual signal (540) from the coded data stream (6) using a context model which depends on (via 520 and 240) as to whether the portion of the enhancement layer signal (360) which the transform coefficient block belongs to, involves inter-layer prediction based on any of the base layer residual signal and the base layer signal, or not; and
- predictively reconstructing (340) the enhancement layer signal (360) using the enhancement layer prediction signal (420).
Q1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a, b, c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360), comprising
- subjecting (220) the reconstructed base layer signal (200) to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- experimentally performing (260), for each coding parameter candidate of a set of coding parameter candidates, a reconstruction or partial reconstruction of the enhancement layer signal (360) using the respective coding parameter candidate to obtain a respective experimental reconstruction or partial reconstruction result;
- determining (260), for each coding parameter candidate, a measure of a difference between the inter-layer prediction signal (380) and the respective experimental reconstruction or partial reconstruction result;
- selecting (260) among the set of coding parameter candidates, depending on the measure for each coding parameter candidate; and
- eventually reconstructing (320, 340, 260) the enhancement layer signal (360) using the selected coding parameter candidate.
Q2) Wherein the coding parameters determined for the enhancement layer relate to intra prediction modes.
Q3) Wherein the coding parameters determined for the enhancement layer relate to motion parameters such as reference pictures or motion vectors.
Q4) Wherein the coding parameters determined for the enhancement layer relate to merge candidates.
Z1) Scalable video decoder configured to
-
- reconstruct (80) a base layer signal (200a, b, c) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360), comprising
- experimentally performing (260), for each coding parameter candidate of a set of coding parameter candidates, a reconstruction or partial reconstruction of the base layer signal (360) using the respective coding parameter candidate to obtain a respective experimental reconstruction or partial reconstruction result;
- determining (260), for each coding parameter candidate, a measure of a difference between the actual decoded base layer reconstruction and the respective experimental reconstruction or partial reconstruction result;
- selecting (260) among the set of coding parameter candidates, depending on the measure for each coding parameter candidate;
- transferring the selected coding parameter candidate to the enhancement layer, where the transferring may include a scaling according to the resolution difference; and
- eventually reconstructing (320, 340, 260) the enhancement layer signal (360) using the transferred coding parameter candidate.
T1) Scalable video decoder configured to
-
- predictively reconstruct (100, 180, 160) a base layer signal by sub-dividing frames into intra-blocks and inter-blocks with using spatial intra prediction for intra-blocks and temporal inter-prediction for inter blocks,
- predictively reconstruct (320, 340, 260) an enhancement layer signal (360) comprising
- locally transferring the block sub-division of blocks of the base layer signal, co-located to a portion of an enhancement layer signal (360), currently to be reconstructed, onto the portion of the enhancement layer signal (360), currently to be reconstructed, so as to obtain sub-portions of the portion of the enhancement layer signal (360), currently to be reconstructed, and associating enhancement layer motion vectors to the sub-portions based on the base layer motion vectors;
- predicting (260), by motion compensated prediction, the portion of the enhancement layer signal (360), currently to be reconstructed, from an already reconstructed portion of the enhancement layer signal (360) using the enhancement layer motion vectors associated with the sub-portions based on the base layer motion vector.
T2) Scalable video decoder configured to
-
- predictively reconstruct (10, 18, 16) a base layer signal using base layer coding parameters spatially varying over the base layer signal;
- reconstruct (32, 34, 26) an enhancement layer signal (36) in units of blocks comprising
- selecting, for a predetermined block of the blocks, a subblock subdivision among a set of possible subblock subdivisions such that the selected subblock subdivision is the coarsest among the set of possible subblock subdivisions subdividing, when transferred onto a co-located portion of the base layer signal, the base layer signal such that within each subblock of the respective subblock subdivision the base layer coding parameters are sufficiently similar to each other;
- predictively reconstruct the predetermined block using the selected subblock subdivision.
U1) Scalable video decoder configured to
-
- predictively reconstruct (100, 180, 160) a base layer signal by sub-dividing frames of the base layer signal into intra-blocks and inter-blocks with using spatial intra prediction for intra-blocks and temporal inter-prediction for inter blocks, and setting prediction parameters associated with the spatial intra and temporal inter prediction, respectively, in units of blocks into which the intra- and inter-blocks are sub-divided;
- predictively reconstruct (320, 340, 260) an enhancement layer signal (360) comprising
- assigning, controlled via prediction mode syntax in the coded data stream, frames of the enhancement signal in coding units into which the frames of the enhancement signal are subdivided to a respective one of a set of prediction modes comprising a spatial intra prediction mode, a temporal inter-prediction mode, and an inter-layer prediction mode,
- predictively reconstructing each coding unit using the respective prediction mode to which the respective coding unit has been assigned with
- for coding units having any of the spatial intra prediction mode and the temporal inter-prediction mode assigned thereto,
- further sub-dividing the coding units into prediction blocks and setting prediction parameters associated with the respective prediction mode to which the respective coding unit has been assigned in units of the prediction blocks; and
- subjecting all prediction blocks to spatial intra prediction using the prediction parameters set, in case of the respective coding unit having the spatial intra prediction assigned thereto, and subjecting all prediction blocks to temporal inter prediction using the prediction parameters set, in case of the respective coding unit having the temporal inter prediction assigned thereto,
- for each coding unit having the inter-layer prediction mode assigned thereto,
- locally transferring the sub-division of the intra and inter-blocks of the base layer signal onto the respective coding unit such that coding units locally overlaying both intra and inter blocks, are sub-divided into at least one prediction block associated with the non-temporal mode and locally coinciding with the intra blocks, and at least one prediction block associated with the temporal inter prediction mode and locally coinciding with the inter blocks;
- subjecting all prediction blocks of the respective coding unit, having the non-temporal prediction mode assigned thereto, to
- spatial intra prediction with using prediction parameters derived from the prediction parameters of the locally coinciding intra blocks, or
- inter layer prediction by
- subjecting (220) the base layer signal to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- predicting (260) the prediction blocks of the respective coding unit, having the non-temporal prediction mode assigned thereto, using the inter-layer prediction signal (380);
- and subjecting all prediction blocks of the respective coding unit, having the temporal inter prediction mode assigned thereto, to temporal inter prediction using the prediction parameters derived from the prediction parameters of the locally coinciding inter blocks.
V1) Scalable video decoder configured to
-
- decode (100) abase layer residual signal (480) of abase layer signal (200) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- decoding a transform coefficient block of transform coefficients representing an enhancement layer signal from the coded data stream, by
- selecting a subblock subdivision among a set of possible subblock subdivisions on the basis of the base layer residual signal or the base layer signal,
- traverse positions of the transform coefficients in units of subblocks into which the transform coefficient block is regularly subdivided in accordance with the selected subblock subdivision such that all positions within one subblock are traversed in an immediately consecutive manner with then proceeding to a next subblock in a subblock order defined among the subblocks,
- for a currently visited subblock,
- decoding from the data stream a syntax element indicating as to whether the current visited subblock has any significant transform coefficient or not;
- if the syntax element indicates that the currently visited subblock does not have any significant transform coefficient, setting the transform coefficients within the currently visited subblock to zero,
- if the syntax element indicates that the currently visited subblock has any significant transform coefficient, decoding from the data stream syntax elements indicating levels of the transform coefficients within the currently visited subblock.
W1) Scalable video decoder configured to
-
- Reconstruct (80) a base layer signal (200) from a coded data stream (6),
- reconstruct (60) an enhancement layer signal (360) comprising
- spatially predicting a block of the enhancement signal by
- subject (220) the reconstructed base layer signal (200a, 200b, 200c) to a resolution or quality refinement to obtain an inter-layer prediction signal (380),
- registering (260)a first line of the inter-layer prediction signal (380), locally overlaying the block, with a second line of an already reconstructed portion of the enhancement layer signal, neighboring the block, wherein the first and second lines are both parallel to a line direction, thereby obtaining a shift value
- filling (260) a line of the block co-located to the first line, with content of the first line, shifted by the shift value.
The line direction may for example be horizontal or vertical, so that that prediction is done row by row or column by column, respectively.
X1) Scalable video decoder configured to
-
- Reconstruct (80) a base layer signal (200) from a coded data stream (6) by
- block-wise prediction,
- a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and
- using an intra prediction parameter for blocks of the base layer signal for which the spatial intra prediction mode has been selected,
- reconstruct (60) an enhancement layer signal (360) from the coded data stream (6) by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an intra prediction parameter for blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, comprising
- checking a neighboring block of the enhancement layer signal, neighboring a current block of the enhancement layer signal, as to whether same has been predicted using the spatial intra prediction mode;
- if yes, appointing the intra prediction parameter of the neighboring block a probably advantageous intra prediction parameter for the current block,
- if not, appointing the intra prediction parameter of a block of the base layer signal, which is co-located to the current block, the probably advantageous intra prediction parameter for the current block,
- determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the probably advantageous intra prediction parameter.
Y1) Scalable video decoder configured to
-
- Reconstruct (80) a base layer signal (200) from a coded data stream (6) by
- block-wise prediction,
- a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and
- using an angular intra prediction parameter for a subset of the blocks of the base layer signal for which the spatial intra prediction mode has been selected,
- reconstruct (60) an enhancement layer signal (360) from the coded data stream (6) by block-wise prediction, a block-wise selection between a spatial intra prediction and a temporal inter-prediction mode, and using an angular intra prediction parameter for a subset of the blocks of the enhancement layer signal for which the spatial intra prediction mode has been selected, comprising
- checking a neighboring block of the enhancement layer signal, neighboring a current block of the enhancement layer signal, as to whether same has been predicted using the spatial intra prediction mode with an angular intra prediction parameter;
- if yes, appointing the angular intra prediction parameter of the neighboring block a probably advantageous intra prediction parameter for the current block,
- if not but a block of the base layer signal, which is co-located to the current block, is coded with an angular intra prediction parameter, appointing the angular intra prediction parameter of a block of the base layer signal, which is co-located to the current block, the probably advantageous intra prediction parameter for the current block,
- determine the intra prediction parameter for the current block based on a syntax element present in the coded data stream for the current block, and the probably advantageous intra prediction parameter.
Although some aspects have been described in the context of an apparatus, it is clear that these aspects also represent a description of the corresponding method, where a block or device corresponds to a method step or a feature of a method step. Analogously, aspects described in the context of a method step also represent a description of a corresponding block or item or feature of a corresponding apparatus. Some or all of the method steps may be executed by (or using) a hardware apparatus, like for example, a microprocessor, a programmable computer or an electronic circuit. In some embodiments, some one or more of the most important method steps may be executed by such an apparatus.
Depending on certain implementation requirements, embodiments of the invention can be implemented in hardware or in software. The implementation can be performed using a digital storage medium, for example a floppy disk, a DVD, a Blu-Ray, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, having electronically readable control signals stored thereon, which cooperate (or are capable of cooperating) with a programmable computer system such that the respective method is performed. Therefore, the digital storage medium may be computer readable.
Some embodiments according to the invention comprise a data carrier having electronically readable control signals, which are capable of cooperating with a programmable computer system, such that one of the methods described herein is performed.
Generally, embodiments of the present invention can be implemented as a computer program product with a program code, the program code being operative for performing one of the methods when the computer program product runs on a computer. The program code may for example be stored on a machine readable carrier.
Other embodiments comprise the computer program for performing one of the methods described herein, stored on a machine readable carrier.
In other words, an embodiment of the inventive method is, therefore, a computer program having a program code for performing one of the methods described herein, when the computer program runs on a computer.
A further embodiment of the inventive methods is, therefore, a data carrier (or a digital storage medium, or a computer-readable medium) comprising, recorded thereon, the computer program for performing one of the methods described herein. The data carrier, the digital storage medium or the recorded medium are typically tangible and/or non-transitionary.
A further embodiment of the inventive method is, therefore, a data stream or a sequence of signals representing the computer program for performing one of the methods described herein. The data stream or the sequence of signals may for example be configured to be transferred via a data communication connection, for example via the Internet.
A further embodiment comprises a processing means, for example a computer, or a programmable logic device, configured to or adapted to perform one of the methods described herein.
A further embodiment comprises a computer having installed thereon the computer program for performing one of the methods described herein.
A further embodiment according to the invention comprises an apparatus or a system configured to transfer (for example, electronically or optically) a computer program for performing one of the methods described herein to a receiver. The receiver may, for example, be a computer, a mobile device, a memory device or the like. The apparatus or system may, for example, comprise a file server for transferring the computer program to the receiver.
In some embodiments, a programmable logic device (for example a field programmable gate array) may be used to perform some or all of the functionalities of the methods described herein. In some embodiments, a field programmable gate array may cooperate with a microprocessor in order to perform one of the methods described herein. Generally, the methods may be performed by any hardware apparatus.
While this invention has been described in terms of several embodiments, there are alterations, permutations, and equivalents which will be apparent to others skilled in the art and which fall within the scope of this invention. It should also be noted that there are many alternative ways of implementing the methods and compositions of the present invention. It is therefore intended that the following appended claims be interpreted as including all such alterations, permutations, and equivalents as fall within the true spirit and scope of the present invention.
Claims
1-20. (canceled)
21. A scalable video decoder, comprising:
- a block-based decoding unit including a processor configured to:
- reconstruct a base layer signal from a data stream; and
- reconstruct an enhancement layer signal of a video from the data stream based on intra prediction using a current intra prediction parameter to decode a current block of the enhancement layer signal, wherein the reconstruct of the enhancement layer signal is based on information about a co-located block of the base layer signal and information about a neighboring block in the enhancement layer signal, wherein the reconstruct of the enhancement layer signal includes:
- checking whether the co-located block of the base layer signal is intra-decoded, wherein the co-located block is co-located relative to the current block of the enhancement layer signal,
- checking whether the neighboring block of the enhancement layer signal is intra-decoded including checking whether a neighboring intra prediction parameter of the neighboring block includes an angular intra prediction parameter, wherein the neighboring block spatially neighbors the current block of the enhancement layer signal,
- responsive to checking the co-located block of the base layer signal, inserting a base-layer intra prediction parameter of the co-located block of the base layer signal into a set of intra prediction parameters for the current block,
- responsive to checking the neighboring block of the enhancement layer signal, inserting the neighboring intra prediction parameter including the angular intra prediction parameter of the neighboring block into the set of intra prediction parameters for the current block,
- extracting, from the data stream, a syntax element identifying at least one of the set of intra prediction parameters, wherein the current intra prediction parameter is determined based on the syntax element present by checking, based on the syntax element, as to whether the current intra prediction parameter is a member of the set of intra prediction parameters,
- determining the current intra prediction parameter for the current block based on the syntax element,
- obtaining an enhancement layer internal prediction of the current block based on the current intra prediction parameter, and
- reconstructing the current block based on a weighted average of the reconstructed base layer signal and the enhancement layer internal prediction of the current block.
22. The scalable video decoder according to claim 21, wherein the block-based decoding unit is configured to perform the checking the neighboring block and inserting in the set with respect to a set of neighboring blocks of the enhancement layer signal, which neighbor the current block.
23. The scalable video decoder according to claim 21, wherein the set of intra prediction parameters for the current block includes a default intra prediction parameter.
24. The scalable video decoder according to claim 21, wherein the base-layer intra prediction parameter is inserted in the set of intra prediction parameters for the current block if the neighboring block of the enhancement layer signal has not been intra-predicted.
25. The scalable video decoder according to claim 21, wherein the base layer signal is reconstructed based on a spatial intra prediction mode or a temporal inter-prediction mode.
26. The scalable video decoder according to claim 21, wherein a default intra prediction parameter is included in the set of intra prediction parameters for the current block if neither the neighboring intra prediction parameter nor the base-layer intra prediction parameter has been selected.
27. The scalable video decoder according to claim 21, wherein if the current intra prediction parameter is a member of the set of intra prediction parameters, then indexing one of the set of intra prediction parameters using the syntax element, and if not, then deriving another one of a set of possible intra prediction modes using the syntax element.
28. A scalable video decoding method, comprising:
- reconstructing a base layer signal from a data stream; and
- reconstructing an enhancement layer signal of a video from the data stream based on intra prediction using a current intra prediction parameter to decode a current block of the enhancement layer signal, wherein the reconstructing of the enhancement layer signal is based on information about a co-located block of the base layer signal and information about a neighboring block in the enhancement layer signal, the reconstructing of the enhancement layer signal including:
- checking whether the co-located block of the base layer signal is intra-decoded, wherein the co-located block is co-located relative to the current block of the enhancement layer signal,
- checking whether the neighboring block of the enhancement layer signal is intra-decoded including checking whether a neighboring intra prediction parameter of the neighboring block includes an angular intra prediction parameter, wherein the neighboring block spatially neighbors the current block of the enhancement layer signal,
- responsive to checking the co-located block of the base layer signal, inserting a base-layer intra prediction parameter of the co-located block of the base layer signal into a set of intra prediction parameters for the current block,
- responsive to checking the neighboring block of the enhancement layer signal, inserting the neighboring intra prediction parameter including the angular intra prediction parameter of the neighboring block into the set of intra prediction parameters for the current block,
- extracting, from the data stream, a syntax element identifying at least one of the set of intra prediction parameters, wherein the current intra prediction parameter is determined based on the syntax element present by checking, based on the syntax element, as to whether the current intra prediction parameter is a member of the set of intra prediction parameters,
- determining the current intra prediction parameter for the current block based on the syntax element,
- obtaining an enhancement layer internal prediction of the current block based on the current intra prediction parameter, and
- reconstructing the current block based on a weighted average of the reconstructed base layer signal and the enhancement layer internal prediction of the current block.
29. The scalable video decoding method according to claim 28, further comprising performing the checking the neighboring block and inserting in the set with respect to a set of neighboring blocks of the enhancement layer signal, which neighbor the current block.
30. The scalable video decoding method according to claim 28, wherein the set of intra prediction parameters for the current block includes a default intra prediction parameter.
31. The scalable video decoding method according to claim 28, wherein the base-layer intra prediction parameter is inserted in the set of intra prediction parameters for the current block if the neighboring block of the enhancement layer signal has not been intra-predicted.
32. A scalable video encoder, comprising:
- a block-based encoding unit including a processor configured to:
- encode a base layer signal of a video into a data stream; and
- encode an enhancement layer signal of the video into the data stream based on intra prediction using a current intra prediction parameter to encode current block of the enhancement layer signal wherein the encode of the enhancement layer signal is based on information about a co-located block of the base layer signal and information about a neighboring block in the enhancement layer signal, the encode of the enhancement layer signal comprising:
- checking whether the co-located block of the base layer signal is intra-encoded, wherein the co-located block is co-located relative to the current block of the enhancement layer signal;
- checking whether the neighboring block of the enhancement layer signal is intra-decoded including checking whether a neighboring intra prediction parameter of the neighboring block includes an angular intra prediction parameter, wherein the neighboring block spatially neighbors the current block of the enhancement layer signal,
- responsive to checking the co-located block of the base layer signal, inserting a base-layer intra prediction parameter of the co-located block of the base layer signal into a set of intra prediction parameters for the current block,
- responsive to checking the neighboring block of the enhancement layer signal, inserting the neighboring intra prediction parameter including the angular intra prediction parameter of the neighboring block into the set of intra prediction parameters for the current block,
- determining the current intra prediction parameter for the current block as at least one of the set of intra prediction parameters, and
- inserting into the data stream a syntax element, which indicates the at least one of the set of intra prediction parameters, wherein the current intra prediction parameter is determined based on the syntax element present by checking, based on the syntax element, as to whether the current intra prediction parameter is a member of the set of intra prediction parameters,
- wherein the current block is to be reconstructed based on a weighted average of a reconstructed base layer signal and an enhancement layer internal prediction of the current block obtained using the current intra prediction parameter.
33. The scalable video encoder according to claim 32, wherein encoding the enhancement layer signal includes checking whether the neighboring block of the enhancement layer signal is intra-decoded, and responsive to checking the neighboring block, inserting the neighboring intra prediction parameter of the neighboring block into the set of intra prediction parameters for the current block.
34. The scalable video encoder according to claim 32, wherein the base-layer intra prediction parameter is inserted in the set of intra prediction parameters for the current block if the neighboring block of the enhancement layer signal has not been intra-predicted.
Type: Application
Filed: Feb 3, 2023
Publication Date: Aug 17, 2023
Inventors: Tobias HINZ (Schordheide), Haricharan LAKSHMAN (Berlin), Jan STEGEMANN (Berlin), Philipp HELLE (Berlin), Mischa SIEKMANN (Berlin), Karsten SUEHRING (Berlin), Detlev MARPE (Berlin), Heiko SCHWARZ (Berlin), Christian BARTNIK (Berlin), Ali Atef Ibrahim KHAIRAT ABDELHAMID (Berlin), Heiner KIRCHHOFFER (Berlin), Thomas WIEGAND (Berlin)
Application Number: 18/164,548
