Selective Subblock Motion Refinement for Consistent Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing DMVR scheme in VVC refines motion vectors (MVs) for each subblock independently, leading to inconsistencies among subblocks, resulting in noticeable subblock boundaries and deteriorating subjective video quality.
Innovation Solution
A method for selectively triggering DMVR by evaluating spatial activity and MV differences between subblocks to determine if DMVR should be applied, preventing inconsistent MVs and subblock boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subblock-based motion refinement (DMVR) is applied to all blocks, then motion vector precision is improved, but subjective video quality deteriorates due to inconsistent MVs and visible subblock boundaries
Solution Approach 1:
The patent applies local quality by selectively enabling DMVR only for specific subblocks based on their characteristics (e.g., motion complexity, texture patterns) rather than uniformly applying it to all subblocks. This ensures that motion refinement is applied where beneficial while avoiding regions where it would create visible boundaries, thus resolving the contradiction between MV precision and subjective quality.
Solution Approach 2:
The patent introduces dynamic decision-making for DMVR application by evaluating block-specific criteria and adapting the refinement strategy accordingly. The system dynamically determines whether to apply DMVR to each block based on local characteristics, rather than using a static all-or-nothing approach, thereby optimizing both precision and quality.
2Measurement precision
If subblock-based motion refinement is applied, then motion vector accuracy is improved, but visible subblock boundaries are created due to inconsistencies among subblocks
Solution Approach 1:
The patent applies local quality by selectively enabling DMVR only for specific subblocks based on their characteristics (e.g., motion complexity, texture patterns) rather than uniformly applying it to all subblocks. This ensures that motion refinement is applied where beneficial while avoiding regions where it would create visible boundaries, thus resolving the contradiction between MV precision and subjective quality.
Solution Approach 2:
The patent merges the motion vector refinement process across subblock boundaries by ensuring consistency checks and coordinated refinement strategies. When DMVR is applied, the system ensures that adjacent subblocks maintain consistent MV relationships, preventing the creation of visible boundaries while preserving accuracy benefits.
3Object-affected harmful factors
If DMVR is selectively triggered based on spatial activity and MV differences, then subjective video quality is improved, but processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-evaluating block characteristics (spatial activity, MV differences) before deciding whether to apply DMVR. This preliminary assessment allows the system to make informed decisions about where refinement is needed, avoiding unnecessary processing in regions where DMVR would not improve quality, thus managing complexity while maintaining quality improvements.
Solution Approach 2:
The patent changes parameters (spatial activity thresholds, MV difference criteria) to optimize the balance between quality improvement and processing complexity. By adjusting these parameters, the system can control the number of blocks subjected to DMVR, thereby managing computational load while preserving subjective quality benefits.
Data Source
AI summary
There is provided a method of encoding a picture. The method comprises selecting a motion vector. MV, for a current block within the picture. The method comprises determining whether a condition is satisfied, wherein determining whether the condition is satisfied comprises: (i) determining if a subblock-based motion refinement is applicable for the current block and (ii) determining if a first subblock within the current block satisfies a criteria. The method comprises, as a result of determining that the condition is satisfied, not using the selected MV for encoding the current block.