Spatial Varying Transform Signaling for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding systems face challenges in efficiently compressing residual blocks with uneven data distribution, leading to increased data size and complexity in signaling transform positions, which affects compression efficiency and rate distortion optimization.
Innovation Solution
The implementation of Spatial Varying Transforms (SVT) with blocks sized half the residual block, limiting positions to vertical or horizontal transforms, and signaling these positions using a candidate position step size and index, reduces the number of candidate positions and signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Spatial Varying Transforms are applied to compress residual blocks, then compression efficiency is improved, but the complexity of signaling transform positions increases
Solution Approach 1:
The patent segments the residual block into multiple candidate positions for SVT blocks, organizing them in a structured grid pattern. This segmentation allows the system to represent transform positions using a limited set of predefined locations rather than arbitrary coordinates, reducing signaling complexity while maintaining compression efficiency.
Solution Approach 2:
The patent changes the parameter space for transform position signaling from continuous or arbitrary coordinates to a discrete set of candidate positions defined by a grid pattern. By transforming the position representation into a indexed form based on candidate positions, the signaling overhead is reduced while preserving the ability to achieve high compression efficiency.
2Productivity
If multiple SVT block positions are allowed, then compression performance is improved, but the number of encoding schemes to test increases
Solution Approach 1:
The patent performs preliminary organization of SVT block positions into a predefined grid of candidate positions before the rate distortion optimization process. This preliminary structuring reduces the search space for RDO by limiting positions to a manageable set of candidates rather than all possible positions, thereby reducing optimization complexity while maintaining compression performance.
Solution Approach 2:
The patent applies SVT blocks to a subset of candidate positions within the residual block rather than requiring evaluation of all possible positions. By selecting a representative set of candidate positions that capture the essential compression benefits, the system achieves good compression performance with reduced computational complexity for rate distortion optimization.
3Manufacturing precision
If SVT block size is reduced to half the residual block, then transform precision is improved, but the number of candidate positions increases
Solution Approach 1:
The patent employs asymmetric SVT block sizes (half the residual block width or height) that are strategically positioned to capture the most significant energy components. This asymmetric design provides transform precision benefits while the positioning strategy keeps the number of meaningful candidate positions manageable through the grid-based candidate position approach.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A video decoding device receives a bitstream including a prediction block and a residual block with coefficients transformed by a Spatial Varying Transform (SVT). The video decoding device determines a type of SVT employed to transform the coefficients in the residual block and determines a position of the SVT relative to the residual block by determining a candidate position step size and a position index for the SVT. The video decoding device applies an inverse transform to the coefficients based on the SVT type and position to create a reconstructed residual block. The video decoding device applies the reconstructed residual block to the prediction block to reconstruct a video block and reconstructs a video sequence for display, the video sequence including a video frame that includes the reconstructed video block.