Apparatuses, systems, and techniques to receive, at one or more processor associated with an image signal processing (ISP) pipeline, a compressed image generated by an image sensor, wherein the compressed image is captured at a first bit-depth associated with the image sensor and is compressed to a second bit-depth that is lower than the first bit-depth, and wherein the ISP is associated with a third bit-depth that is lower than the first bit-depth and higher than the second bit-depth; and decompress the compressed image according to a power curve to generate a partially decompressed image having the third bit-depth, wherein a plurality of regions of the partially decompressed image are decompressed at separate decompression amounts based on a corresponding pixel value of each region of the plurality of regions.

Apparatuses, systems, and techniques to receive, at one or more processor associated with an image signal processing (ISP) pipeline, a compressed image generated by an image sensor, wherein the compressed image is captured at a first bit-depth associated with the image sensor and is compressed to a second bit-depth that is lower than the first bit-depth, and wherein the ISP is associated with a third bit-depth that is lower than the first bit-depth and higher than the second bit-depth; and decompress the compressed image according to a power curve to generate a partially decompressed image having the third bit-depth, wherein a plurality of regions of the partially decompressed image are decompressed at separate decompression amounts based on a corresponding pixel value of each region of the plurality of regions.

A method of multi-scale neural image compression with intra-prediction residuals is performed by at least one processor and includes downsampling an input image, generating a current predicted image, based on a previously-recovered predicted image, and generating a prediction residual based on a difference between the downsampled input image and the generated current predicted image. The method further includes encoding the generated prediction residual, decoding the encoded prediction residual, and generating a currently-recovered predicted image based on an addition of the current predicted image and the decoded prediction residual. The method further includes upsampling the currently-recovered predicted image, generating a scale residual based on a difference between the input image and the upsampled currently-recovered predicted image, and encoding the scale residual.

A method of multi-scale neural image compression with intra-prediction residuals is performed by at least one processor and includes downsampling an input image, generating a current predicted image, based on a previously-recovered predicted image, and generating a prediction residual based on a difference between the downsampled input image and the generated current predicted image. The method further includes encoding the generated prediction residual, decoding the encoded prediction residual, and generating a currently-recovered predicted image based on an addition of the current predicted image and the decoded prediction residual. The method further includes upsampling the currently-recovered predicted image, generating a scale residual based on a difference between the input image and the upsampled currently-recovered predicted image, and encoding the scale residual.

A video decoder, encoder, and corresponding methods for processing video data for an image block and a particular reference picture index to predict the image block are disclosed that utilize adaptive weighting of reference pictures to enhance video compression, where a decoder includes a reference picture weighting factor unit for determining a weighting factor corresponding to the particular reference picture index; an encoder includes a reference picture weighting factor assignor for assigning a weighting factor corresponding to the particular reference picture index; and a method for decoding includes receiving a reference picture index with the data that corresponds to the image block, determining a weighting factor for each received reference picture index, retrieving a reference picture for each index, motion compensating the retrieved reference picture, and multiplying the motion compensated reference picture by the corresponding weighting factor to form a weighted motion compensated reference picture.

A video decoder, encoder, and corresponding methods for processing video data for an image block and a particular reference picture index to predict the image block are disclosed that utilize adaptive weighting of reference pictures to enhance video compression, where a decoder includes a reference picture weighting factor unit for determining a weighting factor corresponding to the particular reference picture index; an encoder includes a reference picture weighting factor assignor for assigning a weighting factor corresponding to the particular reference picture index; and a method for decoding includes receiving a reference picture index with the data that corresponds to the image block, determining a weighting factor for each received reference picture index, retrieving a reference picture for each index, motion compensating the retrieved reference picture, and multiplying the motion compensated reference picture by the corresponding weighting factor to form a weighted motion compensated reference picture.

This application provides a chroma block prediction method and apparatus. The method includes: obtaining a maximum luma value and a minimum luma value based on luma samples corresponding to neighboring samples of a target chroma block, and calculating a first difference between the maximum luma value and the minimum luma value; if the first difference is not equal to 0, processing the first difference based on a quantity of significant bits of the first difference and a first preset bit depth to obtain a second difference; and determining, based on a first chroma value, a second chroma value, and the second difference, an intra prediction model parameter corresponding to the target chroma block, and determining prediction information of the target chroma block based on the intra prediction model parameter and luma reconstruction information corresponding to the target chroma block.

Ultra light models and decision fusion for increasing the speed of intra-prediction are described. Using a machine-learning (ML) model, an ML intra-prediction mode is obtained. A most-probable intra-prediction mode is obtained from amongst available intra-prediction modes for encoding the current block. As an encoding intra-prediction mode, one of the ML intra-prediction mode or the most-probable intra-prediction mode is selected, and the encoding intra-prediction mode is encoded in a compressed bitstream. A current block is encoded using the encoding intra-prediction mode. Selection of the encoding intra-prediction mode is based on relative reliabilities of the ML intra-prediction mode and the most-probable intra-prediction mode.

Ultra light models and decision fusion for increasing the speed of intra-prediction are described. Using a machine-learning (ML) model, an ML intra-prediction mode is obtained. A most-probable intra-prediction mode is obtained from amongst available intra-prediction modes for encoding the current block. As an encoding intra-prediction mode, one of the ML intra-prediction mode or the most-probable intra-prediction mode is selected, and the encoding intra-prediction mode is encoded in a compressed bitstream. A current block is encoded using the encoding intra-prediction mode. Selection of the encoding intra-prediction mode is based on relative reliabilities of the ML intra-prediction mode and the most-probable intra-prediction mode.

An intra prediction method includes obtaining respective intra prediction modes or texture distributions of P reconstructed picture blocks in a surrounding region of a current block; obtaining, based on the respective intra prediction modes or texture distributions of the P reconstructed picture blocks, Q priori candidate intra prediction modes of the current block and Q probability values; obtaining, based on M probability values corresponding to M priori candidate intra prediction modes, M weighting factors corresponding to the M priori candidate intra prediction modes; separately performing intra prediction based on the M priori candidate intra prediction modes to obtain M predicted values; and obtaining a predicted value of the current block based on a weighted summation of the M predicted values and the corresponding M weighting factors.