Abstract: A neural network including at least one neural network layer and an activation function connected to an output of the at least one neural network layer. The activation function is implemented as an approximation function of a mathematically defined real valued non-linear activation function, wherein the approximation function allows for integer-only processing of fixed-point representations of input values of the approximation function.

Abstract: The present disclosure relates to a method of operating a neural network with clipped input data. The method includes defining lower and upper threshold values for integer numbers in data entities of input data for at least one neural network layer. If a value of an integer number in a data entity of the input data is smaller than the defined lower threshold value, the method includes clipping the value of the integer number comprised in the data entity of the input data to the defined lower threshold value. If a value of an integer number in a data entity of the input data is larger than the defined upper threshold value, the method includes clipping the value of the integer number comprised in the data entity of the input data to the defined upper threshold value. Integer overflow of an accumulator register is thereby avoided.

Abstract: A method of operating a neural network based on conditioned weights includes defining integer lower and upper threshold values for values of integer numbers comprised in data entities of input data for the neural network layer. If a value of an integer numbers comprised in a data entity of the input data is smaller than the lower threshold value, the value of the integer number comprised in the data entity of the input data is clipped to the lower threshold value, or if a value of an integer number comprised in a data entity of the input data is larger than the upper threshold value, the value of the integer number comprised in the data entity of the input data is clipped to the upper threshold value. Integer valued weights are determined based on the lower threshold value, the upper threshold value, and a pre-defined accumulator register size, such that integer overflow of the accumulator register can be avoided.

Abstract: Methods and apparatuses are provided for estimating motion vectors of a dense motion field based on subsampled sparse motion field. The sparse motion field includes two or more motion vectors with their respective start positions. For each of the motion vectors, a transformation is derived which transforms the motion vector from its start point into a target point. The transformed motion vectors then contribute to the estimated motion vector on the target position. The contribution of each motion vector is weighted. Such motion estimation may be readily used for video encoding and decoding.

Abstract: A method of decoding of a picture from a video bitstream is disclosed. The video bitstream includes a slice header of a current slice and data representing the current slice. The method comprises: obtaining a parameter used to derive a number of tiles in the current slice from the slice header when a slice address of the current slice is not an address of a last tile in the picture where the current slice located; and reconstructing the current slice using the number of tiles in the current slice and the data representing the current slice.

Abstract: An image processing apparatus processes a color filter mosaic, CFM, image of a scene into a final image of the scene. The image processing apparatus includes processing circuitry configured to implement a neural network. The neural network is configured to process the CFM image into an enhanced CFM image. The processing circuitry is further configured to transform the enhanced CFM image into the final image.

Abstract: The present disclosure relates to image modification such as an image enhancement wherein the processing is at least partially based on neural networks. In particular, the image modification includes a multi-channel processing in which a primary channel is processed separately and secondary channels are processed based on the processed primary channel. The primary channel is processed based on a first spatial frequency transform to obtain a transformed primary channel and the secondary channel is processed based on a second spatial frequency transform to obtain a transformed secondary channel. The transformed primary channel is processed by means of a first neural network to obtain a modified transformed primary channel and the transformed secondary channel is processed based on the transformed primary channel by means of a second neural network to obtain a modified transformed secondary channel.

Abstract: A method for operating a two-dimensional (2D) separable interpolation filter for coding a video, wherein the two-dimensional separable interpolation filter comprises a first and a second one-dimensional (1D) interpolation filter and a temporal buffer of a predetermined buffer size, wherein the method comprises: obtaining filter coefficients for the first one-dimensional interpolation filter of the two-dimensional interpolation filter; comparing a sum of positive filter coefficients of the obtained filter coefficients with a threshold; and upon determining that the sum of positive filter coefficients is larger than the threshold, amending one or more of the filter coefficients to obtain amended filter coefficients, wherein the one or more filter coefficients are amended such that a sum of positive filter coefficients of the amended filter coefficients is not larger than the threshold; applying the amended filter coefficients to samples of the video to obtain a value of a fractional sample position of the video

Abstract: A method of processing a current object is provided. A set of input data tensors representing the current object are inputted into a first neural layer of a transformer based neural network. Based on information about processing the current object, at least one auxiliary data tensor is inputted into the first neural layer or a second neural layer of the transformer based neural network, where the at least one auxiliary data tensor is different from each of the input data tensors of the set of input data tensors and represents at least one auxiliary input. The set of input data tensors are processed by the transformer based neural network using the at least one auxiliary data tensor in order to obtain a set of output data tensors.

Abstract: A motion information storing method comprises: determining a distance between a sample set in the current block and the common boundary; and comparing the distance with a threshold to determine whether third motion information is stored for the sample set, wherein the third motion information is derived by the first motion information and the second motion information.

Abstract: A method of obtaining a chrominance quantization parameter (QP) for chrominance components based on a luminance QP for luminance component, wherein the method is performed by a decoder, comprising parsing a received bitstream to obtain the luminance QP and information of a mapping function (ƒ) which associates a QP index (QPi) to the chrominance QP (QPc); obtaining the QPi based at least in a part on the luminance QP; obtaining the mapping function based on the obtained information; and obtaining a QPc based on the obtained mapping function and the obtained QPi.

Abstract: The present disclosure relates to video encoding and decoding, and in particular to determining motion information for a current block using a history-based motion vector predictor, HMVP, list. The HMVP list is constructed, with said list being an ordered list of N HMVP candidates Hk, k=0, . . . , N?1, which are associated with motion information of N preceding blocks of the frame and precede the current block. Each HMVP candidate has motion information including elements of one or more motion vectors, MVs, one or more reference picture indices corresponding to the MVs, and one or more bi-prediction weight indices. One or more HMVP candidates from the HMVP list are added into a motion information candidate list for the current block; and the motion information is derived based on the motion information candidate list.

Abstract: A method of coding is described. The method can include obtaining a bitstream for a current picture, obtaining a value of a first indicator for the current picture according to the bitstream indicating a slice type, and obtaining a value of a second indicator for the current picture according to the bitstream indicating whether a weighted prediction parameter is present in a picture header or slice header of the bitstream. The method can also include parsing a value of the weighted prediction parameter for a current block of a current slice of the current picture from the bitstream. Furthermore, the method can include predicting the current block according to the value of the weighted prediction parameter.

Abstract: A method for inverse quantization of a current block of a picture or a frame, where the method is performed by a decoder, and includes obtaining, by a decoder, from a bitstream, a joint chrominance residual (JCCR) mode for a current block of a picture or a frame, obtaining one or more quantization parameter (QP) offset values for chrominance components from the bitstream, determining at least one QP value for the current block depending on the JCCR mode, and performing an inverse quantization process on a chrominance component of the current block using the determined QP value.

Abstract: Parallel processing of inter prediction of a current coding block and preceding coding blocks depending on the partition depth of the current coding block is disclosed. An apparatus comprises a processing circuitry configured for: obtaining a value of a parameter for the current coding block, wherein the value of the parameter indicates a partition depth of the current coding block, and performing an inter prediction process for the current coding block; wherein no Motion Vector Predictor candidate comprising a motion vector of a coded block of the picture that is not spatially adjacent to the current coding block for the current coding block is used in the inter prediction process or a merge mode inter prediction process comprised in the inter prediction process, when the value of the parameter for the current coding block is greater than a threshold.

Abstract: A method related to adaptive usage of pixel-based and block-based affine motion compensation and includes performing sub-block-based affine transform prediction on an affine block with a first sub-block size if the first sub-block size is greater than or equal to a first sub-block size threshold. If the first sub-block size is less than the first sub-block size threshold, it is determined whether a pixel-based motion vector field is not applied for performing motion compensation for the affine block. If the motion vector field is not applied, a sub-block-based affine motion prediction is performed with a second sub-block size that is based on the first sub-block size. If the pixel-based motion vector field is applied, motion compensation is performed for a sub-block of the affine block using the pixel-based motion vector field.

Abstract: A method of coding implemented by a decoding device, comprising obtaining a video bitstream; decoding the video bitstream to obtain an initial intra prediction mode value for chroma component of a current coding block; determining whether a ratio between a width for luma component of the current coding block and a width for chroma component of the current coding block is equal to a threshold or not; obtaining a mapped intra prediction mode value for chroma component of the current coding block according to a predefined mapping relationship and the initial intra prediction mode value, when it's determined that the ratio is equal to the threshold; obtaining a prediction sample value for chroma component of the current coding block according to the mapped intra prediction mode value.

Abstract: A method comprising: signaling a set of low-level indicators for a predicted block, wherein each of the indicators of the set enables or disables an intra prediction smoothing mechanism for the predicted block; signaling a first set of high-level indicators on a slice-level or a higher level, wherein each of the high-level indicators of the first set determines signaling for an intra prediction smoothing mechanism which is enabled or disabled by an indicator of the set of low-level indicators; and signaling a second set of high-level indicators, wherein each of the high-level indicators of the second set determines the presence of at least one of the corresponding high-level indicators of the first set, and wherein the values of corresponding high-level indicators of the first set are inferred to be zero when the value signaled by high-level indicator of the second set is zero.

Abstract: The present disclosure provides example computer-implemented methods and decoders of video coding for prediction of a block using an inter prediction mode to obtain predicted samples for the block. One example method includes defining a size of a subblock that is smaller than the size of the block. A position of the subblock within the block is specified. One or more allowed split directions of a triangle partitioning mode (TPM) is determined for the subblock depending on which of the sides of the subblock coincide with the sides of the block. One or more predicted samples for the block is obtained by using TPM partitioning performed for the subblock in accordance with the size of the subblock, the position of the subblock within the block, and TPM split direction of the subblock. The block is reconstructed based on the one or more predicted samples.

Abstract: The present disclosure provides a method of decoding an encoded signal. The method includes receiving at least one bitstream comprising an encoded signal, the signal being entropy encoded with one or more Gaussian mixture models (GMMs), and the at least one bitstream comprising information for obtaining parameters of the one or more GMMs. The method further includes obtaining the GMM parameters based on the information from the at least one bitstream; and entropy decoding the signal using the GMMs with the obtained GMM parameters. The present disclosure further refers to a corresponding encoding method, decoder and encoder.