Template Matching Prediction with Sub-Sampled Search Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding and decoding processes face inefficiencies in the search for prediction blocks, which can be improved by enhancing the speed and efficiency of template matching prediction (TMP) through sub-sampling techniques.
Innovation Solution
Implementing sub-sampled and refinement TMP by determining search areas based on sub-sampling factors, with the option to enable or disable these methods in video data, and performing refinement searches in smaller areas to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching prediction is performed on the full search area, then prediction accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the template matching process into two distinct stages: coarse TMP that operates on a sub-sampled search area to identify candidate prediction blocks, and refinement TMP that operates on a reduced search area around the candidate to improve accuracy. This segmentation allows the system to balance between processing speed and prediction accuracy by performing exhaustive search only on a limited refined area rather than the entire search space.
Solution Approach 2:
The patent implements preliminary action through sub-sampled TMP performed before refinement TMP. The sub-sampled TMP预先 processes a down-sampled version of the search area to identify candidate prediction blocks, which then serve as the basis for the subsequent refinement stage. This preliminary processing reduces the search space for the refinement stage, thereby improving overall encoding/decoding efficiency while maintaining prediction accuracy.
2Productivity
If sub-sampled TMP is performed, then processing speed is improved, but prediction accuracy may deteriorate
Solution Approach 1:
The patent resolves this contradiction through segmentation by separating the template matching into coarse (sub-sampled) and refinement stages. The sub-sampled TMP provides speed improvement by operating on down-sampled data, while the refinement TMP restores accuracy by performing detailed matching on a reduced search area around the candidate prediction block identified in the coarse stage.
Solution Approach 2:
The patent uses an intermediary approach where the candidate prediction block identified through sub-sampled TMP serves as a bridge between the coarse sub-sampled search and the final accurate prediction. This intermediary candidate block allows the system to leverage the speed benefits of sub-sampling while ensuring final accuracy through refinement matching.
3Productivity
If refinement TMP is performed on a smaller search area, then processing efficiency is improved, but the search coverage is reduced
Solution Approach 1:
The patent applies local quality by concentrating computational resources on the most promising regions. The refinement TMP focuses exclusively on a reduced search area around the candidate prediction block identified in the coarse stage, applying high-quality detailed matching only where it is most needed rather than uniformly across the entire search area.
Data Source
AI summary
Devices and techniques for template matching prediction (TMP) with sub-sampling. An example device for video decoding may determine that sub-sampled TMP is enabled for a current block. The device may determine a search area for performing TMP. The device may determine, based on the search area, a sub-sampled search area for performing the sub-sampled TMP. The device may perform the sub-sampled TMP on the sub-sampled search area to determine a candidate prediction block for the current block. The device may decode the current block based on the candidate prediction block.