Merge With MVD Video Coding Through Selective Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity associated with the Merge with MVD Mode (MMVD) technique in video coding is high, leading to increased computational demands and inefficient processing.
Innovation Solution
The method involves determining a set of expanded merge candidates by adding offsets to a base merge MV, restricting steps and directions, and using template matching costs to reorder candidates, while applying these methods within bounding boxes or for small steps to reduce complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full template matching is applied to all expanded merge candidates, then coding performance is improved, but computational complexity increases significantly
Solution Approach 1:
The patent applies template matching selectively only to a subset of expanded merge candidates (e.g., first K candidates) rather than all candidates. This partial action approach maintains sufficient coding performance for the most promising candidates while avoiding the excessive computational burden of evaluating all possible candidates, thus resolving the contradiction between performance and complexity.
Solution Approach 2:
The patent segments the set of expanded merge candidates into different groups (e.g., first K candidates that undergo template matching versus remaining candidates that use alternative selection methods). This segmentation allows the system to apply computationally intensive template matching only where most beneficial while using simpler methods for other candidates, balancing performance and complexity.
2Measurement precision
If all expanded merge candidates are evaluated using template matching, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
The patent implements partial template matching by evaluating only the first K expanded merge candidates using the computationally intensive template matching method, while using faster alternative methods for the remaining candidates. This selective approach maintains high prediction accuracy for the most likely correct candidates while significantly reducing overall processing time compared to evaluating all candidates.
Solution Approach 2:
The patent performs preliminary sorting or ordering of expanded merge candidates based on criteria such as motion vector difference magnitude or spatial proximity before applying template matching. This preliminary action identifies the most promising candidates in advance, allowing template matching to be applied efficiently to the top candidates first, thereby reducing total processing time while maintaining accuracy.
3Productivity
If unrestricted search directions and steps are used, then coding efficiency is improved, but algorithm complexity increases
Solution Approach 1:
The patent applies different search strategies to different regions or types of candidates. For example, it uses restricted search directions and steps for certain candidate groups while allowing more extensive searches for others. This local differentiation allows the algorithm to maintain high coding efficiency for critical regions while reducing overall algorithmic complexity through selective restriction.
Solution Approach 2:
The patent dynamically adjusts search parameters (such as maximum step size, allowed directions, and search radius) based on the specific candidate being evaluated and the current coding context. By changing these parameters adaptively rather than using fixed unrestricted parameters, the algorithm achieves good coding efficiency while controlling algorithmic complexity through parameter-based control.
Data Source
AI summary
A method and apparatus for video coding using MMVD mode are disclosed. According to this method, a base merge MV is determined from a merge list. A set of expanded merge candidates is determined according to a set of steps and a set of directions by adding pairs of the step and direction to the base MV. At least one combination of the set of steps and the set of directions is excluded in a partial set of expanded merge candidates. The partial set of expanded merge candidates are reordered according to template matching costs measured between a template of the current block and a template of a corresponding reference block for in the partial set of expanded merge candidates. The current block is encoded or decoded by using motion information comprising a reordered partial set of expanded merge candidates.