Adaptive Transform Skip in Video Coding for Image Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression methods always perform transform, leading to deteriorated picture quality and reduced compression ratio, especially when the degree of change in spatial pixel values is large, and current technologies cannot skip transform for block sizes greater than 4×4, even if it's inefficient.
Innovation Solution
An encoder and decoder system that determines whether to perform transform based on a maximum transform skip block size, allowing flexible control by comparing the current transform unit size with a predefined Log2MaxTrafoSizeSkip value, and signaling this decision in the bit stream to skip or perform transform accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If transform is always performed in video compression, then compression process is simple, but picture quality deteriorates and compression ratio reduces when spatial pixel value change is large
Solution Approach 1:
The patent applies dynamics by making the transform operation adaptive rather than fixed. The encoder dynamically determines whether to perform transform based on the characteristics of each block (e.g., degree of change in spatial pixel values). When transform is beneficial, it is performed; when it causes quality deterioration, it is skipped. This dynamic adaptation resolves the contradiction between process simplicity and picture quality.
Solution Approach 2:
The patent changes the parameter of transform application from a constant 'always perform' to a conditional parameter based on block characteristics. By modifying this parameter according to the degree of spatial pixel value change, the system optimizes between compression simplicity and picture quality preservation.
2Adaptability or versatility
If transform skip flag is sent for each transform block, then transform skipping is flexible, but signaling overhead increases
Solution Approach 1:
The patent applies local quality by differentiating the treatment of different transform block sizes. Instead of uniformly applying transform skip flags to all blocks, it selectively applies the flag only to 4×4 transform blocks where transform skipping is most beneficial. Larger transform blocks (8×8, 16×16, etc.) do not receive the skip flag, maintaining transform performance for those sizes while reducing overhead.
Solution Approach 2:
The patent applies partial action by implementing transform skip functionality only for a subset of transform blocks (specifically 4×4 blocks) rather than for all block sizes. This partial application reduces the total number of flags needed while still providing transform skipping where it is most effective, thereby reducing overhead.
3Area of stationary object
If smallest transform block size is set to 8×8 or more, then transform block size is larger, but transform cannot be skipped even when efficiency is low
Solution Approach 1:
The patent applies segmentation by dividing the transform block processing into different size categories. It maintains the ability to skip transform for small blocks (4×4) while allowing larger blocks (8×8, 16×16, etc.) to have transform always performed. This segmentation of block sizes enables flexible transform skipping where beneficial while maintaining larger block processing capabilities.
Data Source
AI summary
The present invention relates to a method which can selectively skip a transform to alleviate problems of deterioration in compression rate and image quality which are mainly generated when applying the motion compensation-transform encoding image compression technique which is mainly used currently, to an image having a drastic spatial change in a pixel value such as a computer image, and can skip the transform by comparing a transform size of a current block and the size of a maximum transform skip block with each other, and provides an efficient signaling technique therefor. The method includes the steps of: comparing the size of a current transform unit (TU) with the size of a maximum transform skip block when determining whether to transform a residual signal corresponding to difference between the current block and a prediction block; and determining whether to perform the transform on the basis of the comparison information on the size of the TU and the size of the maximum transform skip block.