Video Block Prediction Using Weighted Reference Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as Intra Block Copy (IBC), face challenges in achieving optimal prediction accuracy due to incomplete consideration of reconstructed block information, particularly in scenarios with linear illumination changes, leading to low prediction accuracy and efficiency.
Innovation Solution
The proposed solution involves determining multiple reference blocks and applying weighted fusion techniques to these blocks based on their respective weight parameters, rather than solely relying on the reconstructed block with the smallest matching cost, to enhance prediction accuracy and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only the reconstructed block with the smallest matching cost is used as the prediction block, then the device complexity is reduced, but the prediction accuracy deteriorates due to incomplete consideration of reconstructed block information
Solution Approach 1:
The patent merges multiple reference blocks (at least two different matching templates) into a single prediction block through weighted fusion. Instead of selecting only one reference block with the smallest matching cost, the method combines information from multiple reference blocks using different weight parameters, thereby improving prediction accuracy while maintaining manageable device complexity through systematic fusion rules.
Solution Approach 2:
The patent introduces weight parameters as a new parameter dimension to control the contribution of different reference blocks. By adjusting the weight parameters associated with each matching template, the system can dynamically change the fusion characteristics to adapt to different content types, thus improving prediction accuracy without significantly increasing device complexity.
2Measurement precision
If multiple reference blocks are used with weighted fusion, then the prediction accuracy is improved, but the device complexity increases due to additional processing operations
Solution Approach 1:
The patent applies local quality by assigning different weight parameters to different reference blocks based on their individual characteristics and matching costs. Each reference block contributes differently to the final prediction block according to its local quality assessment, allowing the system to improve prediction accuracy in critical regions while avoiding unnecessary complexity in regions where simple copying suffices.
Solution Approach 2:
The patent implements partial action by selectively applying weighted fusion only when beneficial, rather than universally processing all blocks with complex fusion algorithms. The method can switch between simple reference block copying and weighted fusion based on content characteristics, thus improving prediction accuracy where needed while maintaining lower device complexity in other scenarios.
3Productivity
If reconstructed block information from different matching templates is fully utilized, then the coding efficiency is improved, but the loss of information is reduced less effectively when relying on a single template
Solution Approach 1:
The patent merges information from multiple matching templates to create a more comprehensive prediction block. By combining the strengths of different templates through weighted fusion, the method reduces information loss that would occur when relying on a single template, thereby improving coding efficiency through better prediction accuracy and reduced residual information.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Disclosed in embodiments of the present application are an encoding/decoding method, a code stream, an encoder, a decoder, and a storage medium. The method comprises: determining N reference blocks of a current block; determining weighted fusion weight parameters of the N reference blocks; performing weighted fusion on the N reference blocks on the basis of the weighted fusion weight parameters of the N reference blocks, and determining a prediction block of the current block; and determining a prediction residual or a reconstruction block of the current block on the basis of the prediction block of the current block. In this way, in the search process of a reconstruction area, reconstruction block information of different matching templates is fully utilized, instead of simply considering reconstruction block information corresponding to a template having a minimum matching cost, so that the prediction accuracy and the coding and decoding efficiency are improved. Furthermore, matching template information is fully utilized to adaptively allocate weight values to reference blocks, it is considered that the prediction importance of different reconstruction block information is different, and the prediction accuracy and the coding and decoding efficiency are further improved.