Decoder-Side Affine Motion Refinement for Non-Translational Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies struggle to accurately model non-translational motion in video data, leading to inefficiencies in compression and quality degradation.
Innovation Solution
Implement decoder-side motion vector refinement using affine models and bilateral matching search to refine motion vectors, enhancing coding accuracy by modeling non-translational motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional translation motion compensation is used, then device complexity is reduced, but coding accuracy deteriorates when modeling non-translational motion
Solution Approach 1:
The patent changes the motion model parameters from simple translation (2 parameters: horizontal and vertical displacement) to affine transformation (6 parameters: 2 translation parameters, 2 rotation parameters, and 2 scaling parameters). This parameter expansion enables accurate modeling of non-translational motion including rotation, scaling, and shearing, directly improving coding accuracy for complex motion patterns while accepting increased computational complexity.
Solution Approach 2:
The patent introduces dynamic motion modeling by allowing different motion types (translation, rotation, scaling, shearing) to be applied to different subblocks within a picture. The system dynamically selects and applies appropriate affine transformation parameters for each subblock based on local motion characteristics, enabling adaptive modeling that improves overall coding accuracy while managing complexity through selective application.
2Measurement precision
If affine models are used for motion compensation, then coding accuracy improves, but bit rate increases
Solution Approach 1:
The patent divides the picture into multiple subblocks and applies affine motion compensation selectively to each subblock rather than uniformly to the entire picture. This segmentation allows the system to apply complex affine models only where non-translational motion is present, improving coding accuracy for those regions while avoiding unnecessary bit rate increase in regions where simple translation suffices.
Solution Approach 2:
The patent applies different motion modeling qualities to different regions by using affine transformations locally in subblocks exhibiting non-translational motion while maintaining simpler models elsewhere. This local quality approach ensures high coding accuracy where needed (rotating or scaling objects) while minimizing overall bit rate consumption by not applying complex models universally.
3Measurement precision
If bilateral matching search is performed, then motion vector precision improves, but processing time increases
Solution Approach 1:
The patent performs preliminary action by conducting bilateral matching search to refine motion vectors before applying the final affine motion compensation. This preliminary refinement of motion vector precision ensures that the subsequent affine transformation is applied to accurately identified corresponding blocks, improving overall coding accuracy while managing processing time through the structured two-stage approach.
Data Source
AI summary
An example method of video coding includes receiving a video bitstream comprising a plurality of blocks. The method also includes deriving a set of motion vectors for a current block of the plurality of blocks and deriving a set of refined motion vectors for the current block using a bilateral matching search and one or more affine models. The method further includes reconstructing the current block using the derived set of refined motion vectors.