Described herein are technologies related to facilitate lossless compression for multi-sample color data of computer graphics that maximizes the apparent quality of pixels while avoiding a corresponding burden on memory and processor bandwidth.

A screen coding method and system based on mass center coincidence. The screen coding method based on mass center coincidence includes: constructing a plurality of coding unit models composed of a combination of a plurality of geometric figures with coincident mass centers, where vertices of the geometric figures do not coincide; and filling in data information to each vertex of the coding unit models according to a method of data information arrangement of a plurality of data combinations to generate a coding unit so as to implement different data lengths of the same coding unit. As such, a data length of the coding unit can be controlled, so that when more data needs to be coded, the overall size of the coding unit does not need to be changed, which greatly improves coding efficiency.

The present invention provides a method for decoding a video signal using a graph-based transform including receiving a generalized graph signal including a graph parameter set; obtaining a graph-based transform kernel of a transform unit based on the graph parameter set and a predetermined penalty function; and decoding the transform unit using the graph-based transform kernel.

In an example, a method of coding video data includes determining, by a video coder and for a block of video data, a palette having a plurality of entries indicating a plurality of respective color values, wherein a first line of the block of video data includes a pixel located adjacent to an edge of the block of video data, and wherein a second line of the block of video data includes a pixel located adjacent to the edge of the block and adjacent to the pixel of the first line. In this example, the method also includes coding, in a scan order, index values that map pixels of the block to entries in the palette, wherein the pixel of the second line immediately follows the pixel of the first line in the scan order.

In an example, a method of coding video data includes determining, by a video coder and for a block of video data, a palette having a plurality of entries indicating a plurality of respective color values, wherein a first line of the block of video data includes a pixel located adjacent to an edge of the block of video data, and wherein a second line of the block of video data includes a pixel located adjacent to the edge of the block and adjacent to the pixel of the first line. In this example, the method also includes coding, in a scan order, index values that map pixels of the block to entries in the palette, wherein the pixel of the second line immediately follows the pixel of the first line in the scan order.

Certain aspects of the present disclosure are directed to methods and apparatus for compressing video content using deep generative models. One example method generally includes receiving video content for compression. The received video content is generally encoded into a latent code space through an auto-encoder, which may be implemented by a first artificial neural network. A compressed version of the encoded video content is generally generated through a trained probabilistic model, which may be implemented by a second artificial neural network, and output for transmission.

Methods and systems are provided for image processing. A plurality of correlation parameters representing degrees of correlation between two or more images of a plurality of images may be produced. An optimized correlation dependency graph may be produced according to the plurality of correlation parameters. The plurality of images may then be delta encoded according to the optimized correlation dependency graph. For example, the optimized correlation dependency graph may be used for performing a correlation encoding operation. The plurality of correlation parameters may be produced, for example, in accordance with one or more correlation metrics associated with the correlation encoding operation.

Methods and systems are provided for image processing. A plurality of correlation parameters representing degrees of correlation between two or more images of a plurality of images may be produced. An optimized correlation dependency graph may be produced according to the plurality of correlation parameters. The plurality of images may then be delta encoded according to the optimized correlation dependency graph. For example, the optimized correlation dependency graph may be used for performing a correlation encoding operation. The plurality of correlation parameters may be produced, for example, in accordance with one or more correlation metrics associated with the correlation encoding operation.

A method of palette coding to apply the palette coding to sub-blocks of a coding unit and to allow each sub-block to use an individual palette table is disclosed. If the current coding block is not partitioned, the palette coding is applied to the current coding block using a first palette. If the current coding block is partitioned into multiple sub-blocks, the palette coding is applied to each sub-block using an individual second palette. Each sub-block may correspond to one prediction unit (PU). In another embodiment of the present invention, a method is disclosed that skips update of palette predictor such as last coded palette table, last coded palette size and palette predictor size associated with the current coding block if the current palette size is smaller than or equal to palette update size.

The present disclosure provides a deep neural network (DNN)-based reconstruction method and apparatus for compressive video sensing (CVS). The method divides a video signal into a key frame and a non-key frame. The key frame is reconstructed by using an existing image reconstruction method. The non-key frame is reconstructed by using a special DNN according to the present disclosure. The neural network includes an adaptive sampling module, a multi-hypothesis prediction module, and a residual reconstruction module. The neural network makes full use of a spatio-temporal correlation of the video signal to sample and reconstruct the video signal. This ensures low time complexity of an algorithm while improving reconstruction quality. Therefore, the method in the present disclosure is applicable to a video sensing system with limited resources on a sampling side and high requirements for reconstruction quality and real-time performance.