Affine Merge Candidate Refinement with Linear Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques using affine merge candidates may not accurately represent the motion of a current block, leading to inefficiencies in compression and distortion, as the affine model of neighboring blocks may not always match the actual motion present in the block to be coded.
Innovation Solution
A video coder refines the affine model for a current block by using a base motion vector field and a guidance motion vector field as inputs to a multivariate linear regression process, producing a refined affine model that determines more accurate affine merge candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If affine merge candidates are derived from the affine model of a neighboring block, then the coding process is simplified and faster, but the accuracy of motion representation deteriorates because the neighboring block may not be spatially close to the current block
Solution Approach 1:
The patent segments the motion representation process into two distinct components: a base motion vector field derived from the neighboring block's affine model (providing computational efficiency), and a guidance motion vector field derived from motion information of adjacent sub-blocks (providing spatial accuracy). This segmentation allows each component to fulfill its specific function optimally while working together to resolve the contradiction between speed and accuracy.
Solution Approach 2:
The patent merges the base motion vector field and the guidance motion vector field through a multivariate linear regression process to produce a refined affine model. This combination integrates the computational advantages of the neighboring block's affine model with the spatial accuracy of adjacent sub-block motion information, thereby achieving both coding efficiency and motion representation accuracy.
2Device complexity
If a simple affine model from a neighboring block is used, then the device complexity is reduced, but the compression efficiency deteriorates due to inaccurate motion representation
Solution Approach 1:
The patent introduces a dynamic refinement mechanism where the affine model is adaptively improved based on local motion characteristics. The multivariate linear regression process dynamically adjusts the base affine model using guidance from adjacent sub-blocks when beneficial, allowing the model complexity to vary according to the actual motion complexity in the current block, thereby achieving better compression efficiency without uniformly increasing device complexity.
Solution Approach 2:
The patent changes the parameters of the affine model by introducing refined motion vector calculations based on multivariate linear regression. This parameter refinement process adjusts the motion vector field to better match the actual motion in the current block, improving compression efficiency while maintaining a relatively simple overall model structure that does not significantly increase device complexity.
3Measurement precision
If the affine model is refined using multivariate linear regression with base and guidance motion vector fields, then the motion representation accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies partial refinement by using the multivariate linear regression process selectively to combine motion vector fields. Rather than completely replacing the simple affine model with a complex refined model, it performs a targeted refinement that focuses computational resources on improving the most critical aspects of motion representation, achieving good accuracy without excessive computational complexity.
Solution Approach 2:
The patent introduces the multivariate linear regression process as an intermediary mechanism that bridges the simple base affine model and the more detailed guidance motion vector field. This intermediary performs the refinement computation in a structured, efficient manner that balances accuracy improvement with computational cost, avoiding the need for overly complex direct modeling approaches.
Data Source
AI summary
A method for coding a block of video data using affine mode includes determining a refined affine model for the current block of video data from a linear regression process using a base motion vector field and a guidance motion vector field as inputs to the linear regression process. The method further includes determining affine merge candidates for the current block using the refined affine model, and coding the current block of video data using the affine merge candidates.