Residual Block Repositioning for Entropy Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video coding, when a transform is skipped for residual blocks in either lossy or lossless coding modes, the positioning of residual data with higher energy values is not optimized for entropy coding, leading to inefficiencies, particularly as these values are not typically positioned at the top left corner of the residual block as assumed by entropy coding algorithms.
Innovation Solution
The technique involves determining whether to reposition residual data by rotating or flipping the residual block based on its prediction mode and size, restricting such repositioning in certain modes and sizes, such as disabling it for inter-prediction and intra block copying modes, and applying rotations like 180 degrees or flipping along diagonals for intra-prediction modes when block sizes are below a threshold, to align higher energy data at the top left corner for improved entropy coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transform is skipped for residual blocks in lossy or lossless coding modes, then coding speed and simplicity are improved, but entropy coding efficiency deteriorates because higher energy residual data is not positioned at the top left corner
Solution Approach 1:
The patent applies repositioning operations (rotation or flipping) to the residual block before entropy coding when transform is skipped. This preliminary action repositions higher energy residual data to the top left corner, ensuring optimal entropy coding efficiency is achieved without requiring transform processing, thus maintaining coding speed while improving compression efficiency.
2Loss of information
If repositioning operations are applied to all residual blocks, then entropy coding efficiency is improved, but device complexity and computational overhead increase
Solution Approach 1:
The patent applies repositioning operations selectively based on local characteristics of the residual block. Specifically, it determines whether rotation or flipping is needed by analyzing the prediction mode and block size, and only applies repositioning when it will actually improve entropy coding efficiency. This selective approach improves compression performance while avoiding unnecessary computational overhead for blocks that don't benefit from repositioning.
3Device complexity
If repositioning is restricted to certain prediction modes and block sizes, then coding complexity is reduced, but entropy coding efficiency is compromised for modes where repositioning could help
Solution Approach 1:
The patent changes the parameters governing repositioning decisions based on prediction mode and block size. For intra-prediction modes with block sizes at or below a threshold (e.g., 8x8), repositioning is enabled to improve entropy coding efficiency. For inter-prediction modes and larger blocks, repositioning is disabled to reduce complexity. This parameter-based control strategy optimizes the balance between compression efficiency and coding complexity for different video content characteristics.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques are described for coding residual data of a prediction residual block in either a lossy or lossless coding mode, in which a transform is skipped or bypassed. The techniques of this disclosure include determining whether to reposition, e.g., rotate or flip, residual data of a residual block prior to coding the residual data of the residual block. For both the lossy and lossless coding modes, a video coding device determines whether to reposition residual data of a residual block based on a prediction mode of the residual block and a size of the residual block. In some examples, the techniques disable repositioning for all residual blocks in the inter-prediction mode and the intra block copying prediction mode, and determine to reposition residual blocks in the intra-prediction mode when the residual blocks have block sizes that are less than or equal to a threshold size.