Template Matching for Video Decoding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In next-generation video codecs, inter-picture prediction using template matching is inefficient due to mismatched positional relationships between blocks, leading to suboptimal processing efficiency.
Innovation Solution
The method involves setting template regions in both the decoded picture and a reference picture, using predetermined vectors to calculate image correlation values, and determining a motion vector to generate a prediction image, with specific conditions such as block size and picture type influencing template region selection to improve processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If blocks are partitioned into subCUs for encoding in next-generation codec, then encoding precision is improved, but template matching processing efficiency deteriorates due to mismatched positional relationships
Solution Approach 1:
The patent segments the template matching process by size: for small subCUs (≤ threshold), it uses adjacent block templates with predetermined positional relationships for efficient processing; for large subCUs (> threshold), it uses traditional template matching for precision. This segmentation resolves the contradiction by applying different strategies to different block sizes.
Solution Approach 2:
The patent applies different template matching approaches based on local characteristics (block size and picture type). Small blocks use adjacent-block templates with fixed positional relationships, while large blocks use traditional templates. This local differentiation optimizes both precision and efficiency for different scenarios.
2Productivity
If traditional template matching is used for all blocks, then processing efficiency is maintained, but encoding precision deteriorates for small blocks
Solution Approach 1:
The patent changes the template selection parameter based on block size: small blocks use adjacent-block templates while large blocks use traditional templates. This parameter change enables optimized processing for different block sizes, improving overall encoding precision without sacrificing processing efficiency.
3Productivity
If adjacent block templates are used for small subCUs, then processing efficiency is improved through parallel processing, but template region availability may be limited
Solution Approach 1:
The patent dynamically selects template regions based on block position and picture type. For non-reference pictures, adjacent blocks are used; for reference pictures, traditional templates are used. This dynamic selection ensures both processing efficiency and template availability across different scenarios.
Data Source
AI summary
A decoding method includes obtaining a rectangular region to be decoded in a picture, setting a first template region in accordance with a position of the rectangular region, setting a second template region corresponding to the first template region in a reference picture, setting third template regions each of which is obtained by moving the second template region using a corresponding vector in the reference picture, calculating image correlation values each of which is obtained between a corresponding one of the third template regions and the first template region, determining a motion vector of the rectangular region, wherein the setting a first template region includes setting a region adjacent to a predetermined block including the rectangular region as the first template region when a size of the rectangular region is a predetermined threshold size or smaller, or when a picture including the rectangular region corresponds to a non-reference picture.