Training a video decoder system may include masking one of at least two sets of video encoding parameters with invalid values to generate an invalid set. The at least two sets of video encoding parameters are provided to one or more neural networks to train them to predict valid values that correspond to values of the invalid set using an iterative training algorithm. Encoding parameters to encode are determined based on a prediction error of the one or more neural networks. Encoding parameters which are determined to be accurately predicted are dropped from the encoded data. A new video stream is encoded without the dropped encoding parameters.

Devices, systems and methods for digital video coding, which includes inter prediction with refinement, are described. An exemplary method of video processing includes determining to use, for a conversion between a current block of a video and a bitstream representation of the video, a first linear optimization model for the conversion using a first coding mode, the first linear optimization model being derived from a second linear optimization model that is used for the conversion using a second coding mode, and performing, based on the determining, the conversion. Another exemplary method of video processing includes determining to use, for a conversion between a current block of a video and a bitstream representation of the video, a gradient value computation algorithm for a bi-directional optical flow tool, and performing, based on the determining, the conversion.

Method and apparatus of video coding for a spherical frame sequence or a cubic frame sequence in a video encoder or decoder are disclosed. According to one method, surrounding blocks for a current block are identified and any surrounding block outside a vertical spherical frame boundary or outside a cubic face boundary of a current cubic face is mapped to a remapped surrounding block. One or more available remapped surrounding blocks for the current block are determined. Mode information reference is generated using mode information associated with said one or more available remapped surrounding blocks. The mode information associated with the current block is then used for encoding or decoding the mode information of the current block. In another method, Intra prediction pixels are determined from the available remapped surrounding blocks. The Intra prediction pixels are used for Intra prediction encoding or decoding of the current block.

A method of video processing includes determining, for a conversion between a current block of a video and a coded representation of the video, that a resolution of a current picture containing the current block and a reference picture used for the conversion are different, and performing the conversion based on the determining such that predicted values of a group of samples of the current block are generated using a horizontal or a vertical interpolation filter.

A compressed dynamic image encoding device is provided, in which a motion vector is generated by searching a reference image for an image area most similar to an image area of a video input signal; a motion-compensated reference image is generated from the motion vector and the reference image; a prediction residual is generated, by subtracting the motion-compensated reference image from the video input signal; the reference image is generated, by adding the motion-compensated reference image and the result of processing performed to the prediction residual; and an encoded video output signal is generated by the processing performed to the prediction residual. The reference image comprises on-screen reference images, located inside a video display screen, and an off-screen reference image located outside the video display screen, and the off-screen reference image is generated based on the positional relationship of plural similar reference images of the on-screen reference images.

A compressed dynamic image encoding device is provided, in which a motion vector is generated by searching a reference image for an image area most similar to an image area of a video input signal; a motion-compensated reference image is generated from the motion vector and the reference image; a prediction residual is generated, by subtracting the motion-compensated reference image from the video input signal; the reference image is generated, by adding the motion-compensated reference image and the result of processing performed to the prediction residual; and an encoded video output signal is generated by the processing performed to the prediction residual. The reference image comprises on-screen reference images, located inside a video display screen, and an off-screen reference image located outside the video display screen, and the off-screen reference image is generated based on the positional relationship of plural similar reference images of the on-screen reference images.

Reconstructing a coded current picture for video decoding including decoding picture partitioning information corresponding to the current picture; determining whether padding is applied to sub-regions of the current picture using the picture partitioning information; based on determining that padding is not applied, decoding the sub-regions without padding the sub-regions; based on determining that padding is applied, determining whether the padding includes wrap-around padding using the picture partitioning information; based on determining that the padding does not include wrap-around padding, applying repetition padding to the sub-regions, and decoding the sub-regions using the repetition padding; based on determining that the padding includes wrap-around padding, applying the wrap-around padding to the sub-regions, and decoding the sub-regions using the wrap-around padding; and reconstructing the current picture based on the decoded sub-regions.

Reconstructing a coded current picture for video decoding including decoding picture partitioning information corresponding to the current picture; determining whether padding is applied to sub-regions of the current picture using the picture partitioning information; based on determining that padding is not applied, decoding the sub-regions without padding the sub-regions; based on determining that padding is applied, determining whether the padding includes wrap-around padding using the picture partitioning information; based on determining that the padding does not include wrap-around padding, applying repetition padding to the sub-regions, and decoding the sub-regions using the repetition padding; based on determining that the padding includes wrap-around padding, applying the wrap-around padding to the sub-regions, and decoding the sub-regions using the wrap-around padding; and reconstructing the current picture based on the decoded sub-regions.

A video encoder according to embodiments is provided. The video encoder is configured for encoding a plurality of pictures of a video by generating an encoded video signal, wherein each of the plurality of pictures includes original picture data. The video encoder includes a data encoder configured for generating the encoded video signal including encoded picture data, wherein the data encoder is configured to encode the plurality of pictures of the video into the encoded picture data. Moreover, the video encoder includes an output interface configured for outputting the encoded picture data of each of the plurality of pictures. Furthermore, a video decoders, systems, methods for encoding and decoding, computer programs and encoded video signals according to embodiments are provided.

A video decoding method according to the present disclosure may include determining whether an affine motion model is applied to a current block or not, performing motion compensation for the current block according to whether the affine motion model is applied or not, determining a value of a first variable and a second variable representing whether a prediction block obtained by the motion compensation will be refined and determining a padding size of the prediction block.