Inter-Prediction Motion Vector Refinement With Segmented Search Space
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving efficient compression ratios with minimal sacrifice in picture quality, particularly in inter-prediction processes, due to limitations in motion vector refinement methods.
Innovation Solution
A method for constructing a search space for motion vector refinement, involving the determination of initial and refining motion vectors through a specific checking order and pattern, including central and neighboring search space positions, to optimize matching costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional motion vector refinement methods are used, then the compression process is simpler and faster, but the compression ratio is lower and picture quality degrades more
Solution Approach 1:
The search space is segmented into multiple regions (first search space and second search space) with different densities. The first search space contains coarser search space positions while the second search space contains finer search space positions. This segmentation allows the algorithm to balance between search thoroughness and computational complexity by performing exhaustive search in critical regions and coarser search in other regions.
Solution Approach 2:
Different regions of the search space are assigned different qualities (densities of search space positions). The first search space positions are distributed at coarser intervals while the second search space positions are distributed at finer intervals. This local quality differentiation optimizes the matching cost calculation by concentrating computational resources where they are most needed while reducing complexity in less critical areas.
2Measurement precision
If a fine search space with many positions is used, then motion vector accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The search space is divided into multiple regions with different densities of search space positions. By segmenting the search space, the algorithm can perform detailed search only in critical regions while using coarser search in other regions, thereby maintaining motion vector accuracy where needed while improving overall compression speed.
Solution Approach 2:
Instead of performing exhaustive fine search across the entire search space, the algorithm performs partial fine search only in specific regions (second search space) while using coarser search in other regions (first search space). This partial action approach achieves sufficient motion vector accuracy without the full computational cost of exhaustive fine search everywhere.
Data Source
AI summary
A non-transitory computer-readable medium for inter-prediction and an encoder are provided. An initial motion vector is obtained for a current block. Search space positions are determined according to the initial motion vector. Matching costs for the search space positions are checked according to a checking order to select a target search space position with a minimal matching cost. A refining motion vector of the current block is determined based on the initial motion vector and the target search space position. A central search space position is determined according to the checking order, and the central search space position is pointed to by the initial motion vector.