Abstract: The invention is made in the technical field of encoding and decoding of bit sequences. The invention proposes a device for compression-encoding an input bit sequence in which Zeroes and Ones occur equally frequent, comprising an XOR gate adapted for receiving pairs of immediately succeeding bits of the input bit sequence as inputs and for outputting further bits of another bit sequence, means for adding a bit to said other bit sequence wherein the added bit is a primary bit which is equal to a primary bit of the input bit sequence or a last bit which is equal to a last bit of the input bit sequence, and means for encoding the other sequence. The XOR gate transforms the input bit sequence into the other bit sequence which comprises more uneven frequencies of occurrence of Zeroes and Ones than the input bit sequence.

Abstract: The invention relates to a method and a device for automatic prediction of a current value using a weighted average of a number of current reference values, wherein the current value is associated with a current pair consisting of a first and a second current data tuple. The method comprises using a set of reference pairs, each reference pair consisting of a first and a second reference data tuple and being associated with a reference value, for selecting the current reference values wherein the first reference data tuples, the first current data tuple and a first metric is used for selecting, determining, for each current reference value, an associated weight using the second reference data tuples of the pair associated with the respective selected reference value, the second current data tuple and a second metric, and using the current reference values and the determined weights for determining the weighted average.

Abstract: When video sequences are obtained from e.g. news, interviews, conferences and sports matches, flashlight often appears in the video due to photographing. When encoding the video sequence, the intensity changes a lot so that the motion estimation can not find a well-matching block in a previous picture. Accordingly, the video encoder will encode in intra mode and thereby generate much more bits than for the neighboring non-flash pictures. According to the invention, not the real flash picture is encoded in the video sequence but an artificial non-flash or de-flashed picture, and some indications and parameters are inserted into the bit stream to model the flash effect. The decoder can regenerate the flash picture by reconstructing the non-flash picture and then adding the flash effect. The number of bits required for coding flash pictures is thereby significantly reduced.

Abstract: Common 2D or 3D mesh models comprise redundancy in the form of symmetries, such as repetitive structures. For complexity reduction, the redundant structures must be detected. An improved method for sampling mesh models comprises sampling the model (710) using an initial sampling step size, detecting (720) a representative of a repeating structure, instances of the repeating structure and a remainder of the model, and sampling (780) the remainder and the representative using a first reduced sampling level according to a first reduced sampling step to size. The method comprises detecting (730) a size of said representative of a repeating structure, a size of the instances of the repeating structure and a size of said remainder, and calculating (740) the first reduced sampling step size based on the size of the instance and the total size of the model. The method can be repeated recursively.

Abstract: The invention is related to a method for compressing images. The proposed method comprises associating perceptual importance parameters with pixels of the image, applying a transform on the image, partitioning the transformed image into code blocks and encoding coefficient bits of a given code block together from a most significant bit plane towards a least significant bit plane, wherein encoding of at least one coefficient is truncated at a truncation bit plane depending on the perceptual importance parameter associated with the pixel to which said code block coefficient corresponds. Truncating encoding of coefficients in dependency on the perceptual importance of the pixel associated with the coefficient allows for truncating coefficients corresponding to perceptually less important pixels at more significant bit planes. So, the overall perceptual quality of the compressed image is enhanced.

Abstract: The invention concerns a method for embedding video annotation data into a coded video stream. The method comprises the step of —encapsulating said video annotation data into a unit, so-called video annotation unit, of the coded video data stream which format corresponds to at least one format used for sending the associated video data, —inserting an identifiable synchronizing code enabling the identification of said video annotation unit into the video data stream.

Abstract: International image or video coding standards uses hybrid coding, wherein a picture is separated into pixel blocks on which predictive coding, transform coding and entropy coding is employed. The transform coding is effective because the prediction error samples are correlated in the frequency domain. However, when the prediction quality is getting better and better, spatial domain coding becomes more effective than transform coding. According to the invention, it is first determined in which corner of a current block the first non-zero amplitude value is located. Based on the related zeros run length value in that block, a pre-defined scan path is selected, i.e. a context-based adaptive scan mode is used.

Abstract: Accurately estimating the image quality of video is still difficult, especially when there is no reference video available. A method for determining a quality measure for a video image comprises selecting a measuring point, such as a MB, determining a context area around the measuring point, calculating a variance of pixel values in the context area, calculating a variance of pixel values in the measuring point, calculating a relation between the two variances of pixel values, and averaging said relations for a plurality of measuring points, wherein a quality measure for a video image is obtained.

Abstract: The present invention provides a salience estimation method for object-based visual attention model. The method comprises steps of segmenting the image into a plurality of objects to be estimated, extracting feature maps for each segmented object, calculating the saliences of each segmented object in a set of circles defined around a center pixel of the object based on the extracted feature maps, and integrating the saliences of each segmented object in all circles in order to achieve an overall salience estimation for each segmented object. The present invention is much more human vision inosculated and of low computing complexity.

Abstract: In known video coding standards there are basically three types of pictures: I, P and B. Video sequences are usually coded in a group of picture structure wherein several P or B pictures are coded following one I picture. However, this GOP structure has some disadvantages with respect to error resilience and storage medium recording. If a P picture is lost e.g. due to transmission channel error, then the subsequent P pictures can not be reconstructed correctly, and the error will propagate temporally and cause unpleasing artifacts. This GOP structure is designed for forward play only and makes complicated a reverse play operation. According to the invention, a reversible GOP structure is used for the video encoding and decoding. The RGOP structure contains both, a forward encoding chain and a backward encoding chain. Each picture in the RGOP structure is assigned to only one of these chains and the video pictures of adjacent RGOP chains are interleaved.

Abstract: A method for encoding vertex orientations, also known as normal components, of a 3D mesh model comprises a first clustering of the normal components, determining for each of the first clusters a sphere sector to which most of its elements belong, mapping normal components to a predefined sector, re-clustering the normal components in the predefined sector into second clusters, determining predictors for the second clusters and encoding a normal component by its residual, a reference to its predictor and data indicating which of said mirror mapping operations were performed. A sphere sector is obtained by dividing a sphere into m equal spherical segments and dividing each of said spherical segments into n equal sectors. The first clustering for the sector mapping and the second clustering for predictive coding result in an improved compression ratio.

Abstract: High-fidelity video coding nowadays tends to adopt the RGB coding for directly reducing RGB inter-color redundancy without color space conversion. A known method uses a blockbased inter-colour compensation algorithm for removing inter-colour redundancy on RGB colour planes by defining the correlation between colour planes within each coding block as a linear model of slope and offset parameters, codes the base plane block and then predicts the other two colour blocks from the base plane block. However, for most coding blocks the texture within a component block varies quite much. A new method for improved video coding uses adaptive segmentation of the base colour component block, corresponding segmentation of the other colour component blocks and individual slope and offset parameters for predicting each segment.

Abstract: The invention is related to encoding an image block of an image using a partitioned block transform. The inventors recognized that applying a texture-pattern associated invertible mapping to the pixels of a first partition, said first partition resulting from partitioning said image block according to a current texture pattern with which said texture-pattern associated invertible mapping is associated, allows for limiting the maximum number of required first 1-D transforms to not exceeding a number of columns in the image block as well as limiting the maximum number of required second 1-D transforms to not exceeding a number of rows in the image block, also. Achieving limitation of maximum required 1-D transforms enables more efficient implementation on hardware and improves encoding performance but also allows for further partitions according to texture patterns which comprise at least one of multiple strips, texture patterns with highly unsymmetrical pixel distribution and non-directional texture patterns.

Abstract: As to enhance the rendering while minimizing needed computation time, the method comprises the steps of determining noise information, said noise information being representative of noise level estimated for at least a part of said image; the noise information being determined for the achromatic component of the image; sampling said part of the image according to said noise information; and rendering said at least a part of the image according to said sampling.

Abstract: For most large 3D engineering models, the instance positions of repeating instances of connected components show significant multiple spatial aggregation. The invention uses several KD-trees, each for one cluster of points which are spatially aggregated. The multiple KD-trees generate a relatively short data stream, and thus improve the total compression ratio. A method for encoding points of a 3D mesh model comprises steps of determining that the mesh model comprises repeating instances of a connected component, and determining for each repeating instance at least one reference point, clustering the reference points of the repeating instances into one or more clusters, and encoding the clustered reference points using KD-tree coding, wherein for each cluster a separate KD-tree is generated.

Abstract: An algorithm for efficiently compressing floating-point data in 3D meshes is disclosed. 3D meshes are represented by topology data, geometry data and property data. Geometry data specify vertex locations and are usually represented by floating-point coordinates. While geometry data are usually compressed by quantization, prediction and entropy coding, the present invention uses no prediction. A floating-point number consists of mantissa and exponent, and normally the exponent, sign and mantissa are compressed separately. A method for encoding floating-point formatted data comprises determining if a current floating-point value was previously stored in a memory, storing the current value in the memory if it was not previously stored in the memory, and encoding it. Otherwise, if the current floating-point value was previously stored in a memory, the storage position of the value within the memory is determined and a reference pointing to the storage position is encoded.

Abstract: Optimised bit allocation is important in video compression to increase the coding efficiency, i.e. to make optimum use of the available data rate. In view of the human visual system, a human usually pays more attention to some part of a picture rather than to other parts of that picture. Therefore the bit allocation should be optimised for different-attention picture areas (GOBi). The inventive distortion-driven bit allocation scheme allocates the coding/decoding error distortion to picture areas consistently with the human visual system, and satisfies the constraint of bit rate as well. The invention uses a distortion/bitrate/rhoquantization parameter histogram analysis. Based on corresponding tables (DGOBi[QPn], RGOBi[QPn] and ?GOBi[QPn]), the relationships between quantization parameter, rate, distortion and percentage of non-zero coefficients for the different-attention areas are determined (PREALUTI, DISALL, RALL).

Abstract: The frame following a scene cut is usually coded as an I picture. In CBR encoding, the encoder will try to keep the bit rate constant, which will often cause serious picture quality degradation at scene changes. In VBR encoding, more bits will be allocated to the first frame of the new scene and the bit rate will increase significantly for a short time. Therefore subsequent frames must be coded in ‘skipped’ mode, which will often cause jerk artifacts. According to the invention, in each frame belonging to a scene change period, areas are determined that have different human attention levels. In the frames (n?1, n?2, n?3) located prior to the first new scene frame, to the areas having a lower attention level less bits are assigned than in the default encoding, and in the frames (n, n+1, n+2) located at and after the scene cut the thus saved bits are additionally assigned to the areas having a higher attention level.

Abstract: 3D models of the engineering class usually have a large number of connected components, with small numbers of large triangles, often with arbitrary connectivity. To enable compact storage and fast transmission of large 3D mesh models, an efficient compression strategy specially designed for 3D mesh models is provide. A method for encoding a 3D mesh model comprises determining and clustering repeating components, normalizing the components, wherein scaling factors are clustered and orientation axes are clustered, encoding the connected components using references to the clusters, and entropy encoding the connected components.

Abstract: A method for compressing a symbol sequence, wherein each symbol may have one out of three or more possible symbol values, said method comprises the steps of modifying the symbol sequence by replacing each run pair, which contains a first run of symbols of a most frequent symbol value and a second run of symbols of a second most frequent symbol value, by a further symbol value not comprised in the three or more possible symbol values, generating a binary sequence comprising all replaced run pairs and compression encoding the binary sequence and the modified symbol sequence.