Irregular Image Block Partitioning for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding and decoding technologies face difficulties in accurately partitioning image blocks with rich texture characteristics, as rectangular sub-blocks fail to adapt to the varying pixel value distributions.
Innovation Solution
The method involves irregularly partitioning image blocks into non-rectangular sub-blocks based on position parameters, comparing these sub-blocks with reference sub-blocks to calculate residuals, and encoding the position parameter differences for efficient coding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If rectangular sub-blocks are used for partitioning image blocks, then the coding and decoding process is simple, but the prediction accuracy for image blocks with rich texture characteristics deteriorates
Solution Approach 1:
The image block is segmented into multiple sub-blocks using irregular partitioning modes (triangular, pentagonal, hexagonal shapes) instead of uniform rectangular division. This segmentation allows each sub-block to better conform to the actual texture boundaries and pixel value distributions within the image block, thereby improving prediction accuracy while maintaining manageable complexity through predefined partitioning patterns.
Solution Approach 2:
Different sub-blocks within the same image block are allowed to have different shapes and sizes based on local texture characteristics. The partitioning adapts to local variations in pixel value distribution, applying appropriate irregular shapes to different regions. This local adaptation enables more accurate prediction in textured areas while preserving simplicity in smooth regions.
2Measurement precision
If irregular partitioning modes are used to partition image blocks, then the prediction accuracy for pixel values improves, but the device complexity increases
Solution Approach 1:
The partitioning structure is made dynamic by selecting from multiple predefined irregular partitioning modes based on the actual content characteristics of each image block. Rather than using a fixed rectangular partitioning scheme, the system dynamically chooses the most appropriate partitioning mode (triangular, pentagonal, hexagonal, etc.) for each block, allowing the structure to adapt to content requirements while maintaining implementation feasibility through a有限的 set of predefined patterns.
Solution Approach 2:
The partitioning approach changes key parameters such as sub-block shape, size, and arrangement based on the texture characteristics and pixel value distributions of the image block. By adjusting these partitioning parameters dynamically, the system achieves higher prediction accuracy for complex textures while managing device complexity through parameter-based adaptation rather than structural overhauls.
3Loss of substance
If position parameter differences are transmitted in the code stream, then the code stream size is reduced, but the decoding complexity increases
Solution Approach 1:
The method extracts and transmits only the position parameter differences (offsets) rather than complete position parameters for each sub-block. By taking out only the essential differential information needed to reconstruct the partitioning structure, the code stream size is significantly reduced. The decoder can reconstruct the full position parameters by adding these differences to reference values, thereby achieving compression without excessive decoding complexity.
Solution Approach 2:
The encoding process performs preliminary calculation of position parameter differences before transmission. By pre-computing these differences and organizing them in a structured format, the transmitted data requires minimal processing during decoding. The preliminary organization of differential data enables the decoder to efficiently reconstruct the partitioning structure with reduced computational burden.
Data Source
AI summary
A coding method includes: partitioning an image block to be decoded irregularly according to an irregular partitioning mode and a position parameter to obtain at least two different sub-blocks; comparing the sub-blocks with a reference sub-block to obtain a residual of a pixel value of the image block to be coded; obtaining a reference position parameter according to a surrounding block of the image block to be coded, and subtracting the reference position parameter from the position parameter of the image block to be coded to obtain a position parameter difference of the image block to be coded; quantizing the residual, and performing entropy coding on a result of the quantization, the motion vector information, and the position parameter difference; and writing the entropy coding result and partitioning mode information of the image block into a code stream.