Intra Template Matching Prediction Using Ordered Reference Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital video compression standards face challenges in efficiently reducing video data volume to alleviate bandwidth and traffic pressure, particularly in high-definition video transmission, due to limitations in intra prediction methods for spatial redundancy and inter prediction for temporal redundancy.
Innovation Solution
The implementation of an intra template matching prediction method that constructs a candidate list for intra template matching prediction by determining a search range, searching for reference block templates, calculating differences, and ordering reference blocks based on similarity, to enhance encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional intra prediction methods are used to eliminate spatial redundancy, then encoding complexity is reduced, but compression performance is insufficient for high-definition video
Solution Approach 1:
The patent segments the prediction process by dividing the current block into multiple sub-blocks and constructing separate candidate lists for each sub-block. This segmentation allows the algorithm to focus computational resources on local variations within different regions, improving compression performance while managing encoding complexity through localized processing rather than global optimization.
Solution Approach 2:
The patent performs preliminary action by pre-construction of candidate lists containing potential reference blocks before the actual prediction process. These candidate lists are built using template matching on previously decoded blocks, preparing potential prediction sources in advance. This preliminary preparation reduces the computational burden during the main encoding phase, balancing compression performance with encoding complexity.
2Measurement precision
If template matching is performed over a large search range to improve prediction accuracy, then compression performance increases, but encoding time and complexity increase
Solution Approach 1:
The patent divides the large search range into multiple smaller search regions corresponding to different sub-blocks. Each sub-block performs template matching only within its localized search region rather than scanning the entire large range. This segmentation maintains prediction accuracy for each local region while dramatically reducing the total number of comparisons required, thus decreasing encoding time.
Solution Approach 2:
The patent applies local quality by allowing different search ranges and candidate list constructions for different sub-blocks based on their local characteristics. Each sub-block can have its own optimized search strategy tailored to its specific content and position, improving prediction accuracy locally while avoiding the computational cost of uniform large-range searching across the entire block.
3Measurement precision
If multiple candidate reference blocks are evaluated to improve prediction quality, then compression performance improves, but encoding overhead increases
Solution Approach 1:
The patent extracts only the most promising candidate reference blocks from the search results and places them into a limited-size candidate list. Rather than evaluating all possible reference blocks, the algorithm selectively extracts the top candidates based on template matching criteria. This extraction approach maintains prediction quality by keeping the best options while reducing encoding overhead by limiting the number of candidates that need to be signaled and processed.
Solution Approach 2:
The patent applies partial action by evaluating a limited subset of candidate reference blocks rather than exhaustively evaluating all possible candidates. The candidate list contains a predetermined number of candidates (e.g., 2-5) which is sufficient to achieve good prediction quality without requiring exhaustive search. This partial evaluation approach balances prediction quality with encoding overhead by doing just enough work to achieve acceptable results.
Data Source
Figure 1A~1B
Figure 1C~2A
Figure 2B~3
AI summary
An intra template matching prediction (intraTMP) method and apparatus, a video coding method and apparatus, a video decoding method and apparatus, and a system. The intraTMP method comprises: when intra prediction is performed on a current block, searching for reference block templates in a reconstruction area for a current image, calculating the difference between the found reference block templates and a current block template, constructing a candidate list of intraTMP according to the difference, and determining N reference blocks in the candidate list and the sequence of the N reference blocks, wherein N>2.