TIMD Intra Prediction Weighting for Complex Texture Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of intra prediction in H.266/VVC's Template-based Intra Mode Derivation (TIMD) manner needs improvement due to varying prediction effects across different positions within a block, especially for complex textures.
Innovation Solution
The current block is partitioned into sub-templates, and weights for each prediction mode are determined based on the performance of these sub-templates, allowing for tailored intra prediction values by considering the specific texture and spatial correlation of different block parts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Template-based Intra Mode Derivation (TIMD) is used for intra prediction, then coding efficiency is improved, but prediction accuracy deteriorates for complex textures due to varying prediction effects across different positions within a block
Solution Approach 1:
The current block is divided into multiple sub-blocks, and separate prediction modes are selected for each sub-block based on template matching. This segmentation allows different prediction modes to be applied to different regions with varying texture characteristics, thereby improving prediction accuracy for complex textures while maintaining coding efficiency through selective mode application.
Solution Approach 2:
The patent applies different prediction modes to different sub-blocks based on local template matching results. Each sub-block receives a prediction mode optimized for its specific local characteristics, rather than applying a single uniform mode to the entire block. This local quality approach resolves the contradiction by adapting prediction accuracy to local texture variations while maintaining overall coding efficiency.
2Measurement precision
If separate prediction modes are selected for different sub-blocks, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The block is segmented into sub-blocks with template matching performed on each. This segmentation enables accurate prediction by capturing local variations, while the systematic approach to segmentation and mode selection keeps computational complexity manageable through reuse of template data and efficient mode decision algorithms.
Solution Approach 2:
The patent applies template matching and mode selection to only the necessary sub-blocks rather than processing the entire block uniformly. This partial action approach improves prediction accuracy where needed (in complex texture regions) while avoiding unnecessary computational overhead in regions where simpler prediction suffices, thus balancing accuracy and complexity.
Data Source
AI summary
A method for intra prediction, applied to an encoder, the method includes: a prediction mode parameter of a current block is acquired, herein the prediction mode parameter indicates that an intra prediction value of the current block is determined by using a Template-based Intra Mode Derived (TIMD) mode; a template of the current block is determined, herein the template includes at least two sub-templates; a weight of each of at least two prediction modes on a first unit of the current block is determined according to respective costs of the at least two prediction modes on a first sub-template, herein the at least two sub-templates include the first sub-template; and the intra prediction value of the current block is determined according to the respective weights of the at least two prediction modes on the first unit.