Template Matching Prediction Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Template matching prediction in video encoding and decoding leads to high complexity at the decoder due to the need for sub-pel interpolation and storage of interpolated frames, despite only a few sub-pel positions being necessary, resulting in increased memory and computation requirements.
Innovation Solution
Constraining template matching prediction using criteria such as limiting search precision to full-pel, reducing the number of searching points, and employing a simpler sub-pel interpolation scheme to reduce complexity, while maintaining coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching prediction uses sub-pel interpolation to improve coding efficiency, then coding efficiency is improved, but decoder complexity and memory consumption increase
Solution Approach 1:
The patent changes the precision parameter from sub-pel to full-pel for template matching prediction. This parameter change resolves the contradiction by maintaining coding efficiency through template matching while eliminating the need for complex sub-pel interpolation at the decoder, thus reducing decoder complexity
Solution Approach 2:
The patent extracts and removes the sub-pel interpolation step from the template matching prediction process at the decoder. By taking out this complex operation, the patent maintains the essential template matching functionality while eliminating the source of decoder complexity and memory consumption
2Measurement precision
If template matching prediction stores interpolated frames at sub-pel precision to improve accuracy, then prediction accuracy is improved, but memory consumption increases
Solution Approach 1:
The patent changes the storage precision parameter from sub-pel to full-pel for reference frames used in template matching prediction. This resolves the contradiction by maintaining sufficient prediction accuracy while significantly reducing the memory required to store interpolated frames
Solution Approach 2:
The patent applies partial action by performing sub-pel interpolation only at the encoder for generating prediction candidates, while the decoder operates at full-pel precision. This partial application of sub-pel processing maintains prediction accuracy where needed while avoiding excessive memory consumption at the decoder
3Measurement precision
If template matching prediction performs comprehensive search at all sub-pel positions to improve precision, then matching precision is improved, but computational complexity increases
Solution Approach 1:
The patent changes the search precision parameter from sub-pel to full-pel for the actual template matching search at the decoder. This resolves the contradiction by maintaining matching precision through encoder-side sub-pel processing while reducing computational complexity at the decoder by eliminating exhaustive sub-pel searches
Solution Approach 2:
The patent performs sub-pel interpolation and candidate generation as preliminary actions at the encoder before transmission. This preliminary processing maintains matching precision while shifting the computational burden from the decoder to the encoder, reducing decoder computational complexity
Data Source
AI summary
There are provided methods and apparatus for reduced complexity template matching prediction for video encoding and decoding. The encoding method encodes at least an image block in a picture using template matching prediction (615, 620, 640, 641, 642), wherein the template matching prediction is selectably constrained using one or more constraining criterion that reduces a complexity of performing the template matching prediction (615, 625, 640).