Motion Candidate Derivation via Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face inefficiencies in deriving motion information, particularly in inter prediction, where signaling motion vectors can be cumbersome and lead to increased bitstream size, and existing methods do not effectively optimize motion vector derivation for improved compression efficiency.
Innovation Solution
The method involves determining a set of motion candidates for a current video block, refining them through local searches based on a template matching cost rule, and selecting a target motion candidate for conversion, which adapts to the coding mode and reference templates, thereby optimizing motion vector derivation and reducing bitstream size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion information is explicitly signaled in the bitstream, then motion vector accuracy is improved, but bitstream size increases
Solution Approach 1:
The system enables the decoder to autonomously derive motion vector candidates by performing template matching and local searches using reference picture data already available at the decoder side, eliminating the need for the encoder to explicitly signal motion information while maintaining accurate motion compensation
Solution Approach 2:
The patent introduces an intermediary derivation process where motion candidates are obtained through template matching against reference blocks and refined via local searches, serving as a bridge between the encoded bitstream and the final motion compensated prediction without requiring direct motion vector signaling
2Measurement precision
If template matching with local search is performed for each motion candidate, then motion vector precision is improved, but computational complexity increases
Solution Approach 1:
The system performs preliminary template matching to identify candidate motion vectors before conducting more computationally intensive local searches, pre-filtering the search space to reduce the number of candidates that require refinement and thereby lowering overall computational complexity
Solution Approach 2:
The motion vector derivation process is segmented into distinct stages: initial template matching to generate candidate vectors, followed by optional local search refinement for selected candidates, allowing the system to apply different levels of computational effort to different candidates based on their likelihood of being optimal
Data Source
AI summary
Embodiments of the present disclosure provide methods, apparatuses and computer storage media for video processing. One example method comprises determining, during a conversion between a current video block of a video and a bitstream of the video, a set of motion candidates for the current video block; determining, for each motion candidate in the set of motion candidates, a refined motion candidate by performing a local search around the each motion candidate based on a template matching cost rule; determining, from a set of refined motion candidates generated for the set of motion candidates, a target motion candidate for the current video block; and performing the conversion based on the target motion candidate.