Intra Template Matching Prediction With Fractional-Pel Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in intra template matching prediction, leading to suboptimal compression and quality degradation in digital video transmission and storage.
Innovation Solution
Implementing a method and apparatus for video decoding that utilizes intra template matching prediction (TMP) by obtaining a current template, an optimal integer predicted template, and fractional-pel predicted templates to enhance coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If intra template matching prediction is implemented using only integer-pel precision, then device complexity is reduced, but manufacturing precision (prediction accuracy) deteriorates
Solution Approach 1:
The prediction process is segmented into two distinct stages: integer-pel template matching to identify candidate positions, followed by fractional-pel interpolation to refine the prediction. This segmentation allows the system to benefit from both low-complexity integer operations and high-precision fractional operations without combining them into a single complex process.
Solution Approach 2:
Integer-pel template matching is performed as a preliminary action to identify candidate positions before conducting fractional-pel interpolation. This preliminary step narrows down the search space and provides initial predictions that are then refined, avoiding the need to perform complex fractional operations across the entire search range.
2Manufacturing precision
If fractional-pel precision is used throughout the entire prediction process, then prediction accuracy is improved, but computational cost increases
Solution Approach 1:
The prediction process is divided into two stages with different precision levels: integer-pel template matching followed by fractional-pel interpolation. This segmentation allows the system to perform most computations at lower precision (faster) and only refine specific candidates at higher precision (slower but more accurate).
Solution Approach 2:
Fractional-pel interpolation is applied partially - only to the best candidate positions identified by integer-pel matching, rather than to all positions. This partial application of the more computationally intensive operation maintains accuracy where needed while preserving overall processing speed.
3Manufacturing precision
If multiple candidate templates are evaluated with high precision, then prediction quality is improved, but processing time increases
Solution Approach 1:
Integer-pel template matching serves as a preliminary filtering stage that quickly identifies the best candidate positions. By performing this coarse evaluation first, the system avoids applying time-consuming fractional-pel interpolation to all candidates, instead only to the most promising ones.
Solution Approach 2:
High-precision fractional-pel interpolation is applied to only the top candidate positions identified by integer-pel matching, rather than to all candidates. This selective application of excessive precision where most needed optimizes the trade-off between prediction quality and processing time.
Data Source
AI summary
Methods for video decoding and encoding, apparatuses and non-transitory computer-readable storage media thereof are provided. In one method for video decoding, a decoder may obtain a current template of a current coding unit (CU) coded with Intra template matching prediction (TMP) mode. Additionally, the decoder may obtain an optimal integer predicted template. Furthermore, the decoder may obtain fractional-pel predicted templates. Moreover, the decoder may obtain a final prediction for the current CU based on a distance between the current template and the optimal integer predicted template and the fractional-pel predicted templates.