Reference Block Weighting for Accurate Inter Prediction Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies face challenges in achieving accurate prediction, reducing residual bits, and enhancing coding efficiency, particularly in inter prediction methods where motion vector signaling and residual blocks are involved.
Innovation Solution
The proposed method involves determining weight information for combining multiple reference blocks based on their respective reference templates to generate a prediction block, which is then used to encode or decode current blocks, thereby improving prediction performance and reducing the data required for inter prediction mode signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single reference block is used for prediction, then the signaling overhead is low, but the prediction accuracy is insufficient
Solution Approach 1:
The patent combines multiple reference blocks from different reference images into a single prediction block by determining weight information for each reference block based on reference templates. This merging approach improves prediction accuracy by utilizing information from multiple references while avoiding the need to signal multiple separate prediction modes, thus reducing signaling overhead.
Solution Approach 2:
The patent creates a composite prediction block by combining multiple reference blocks with different weightings. This composite approach is analogous to using composite materials, where the final prediction is formed by combining multiple source blocks with optimized weights based on template matching, achieving superior prediction accuracy compared to using a single reference block.
2Measurement precision
If multiple reference blocks are combined with weight information, then the prediction accuracy improves, but the amount of data to signal increases
Solution Approach 1:
The patent employs self-service by deriving the weight information automatically from the reference templates themselves. The template matching process inherently provides the weighting factors needed for combining reference blocks, eliminating the need for separate weight signaling. The system serves itself by extracting all necessary information from the existing reference data.
Solution Approach 2:
The patent changes the parameter representation by using reference template similarity metrics directly as weight information. Instead of signaling separate weight parameters, the system transforms the template matching results into usable weight values, achieving accurate multi-reference prediction without increasing the signaled data amount.
3Quantity of substance
If conventional inter prediction is used, then the coding process is simple, but the residual bits are excessive
Solution Approach 1:
The patent performs preliminary action by determining weight information and generating the optimal prediction block before the residual calculation stage. By pre-computing the weighted combination of reference blocks based on template matching, the system creates a more accurate prediction that closely matches the current block, thereby minimizing the residual information that needs to be encoded.
Data Source
AI summary
A method for image decoding includes determining a first reference block in a first reference image and a second reference block in a second reference image corresponding to a current block; determining weight information for combining the first reference block and the second reference block based on a current reference template of the current block, a first reference template of the first reference block, and a second reference template of the second reference block; generating a prediction block of the current block by combining the first reference block and the second reference block based on the weight information; and reconstructing the current block based on the prediction block.