Affine Motion Compensation for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC, face limitations in accurately predicting complex motions like zoom, rotation, and perspective within video frames, as they primarily use translation motion models, leading to inefficiencies in compression and quality retention.
Innovation Solution
The implementation of affine and non-affine motion compensation techniques, where motion vectors are updated and used across subsequent blocks, allowing for more precise prediction and improved compression efficiency by utilizing affine models derived from neighboring blocks and adaptive motion vector prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If translation motion models are used for video coding, then device complexity is reduced and ease of operation is improved, but manufacturing precision (motion prediction accuracy) deteriorates for complex motions like zoom, rotation, and perspective
Solution Approach 1:
The current block is divided into multiple sub-blocks, each with its own motion vector derived from affine parameters. This segmentation allows different regions to capture different motion characteristics, improving overall motion prediction accuracy for complex motions while keeping the computational complexity manageable through localized processing
Solution Approach 2:
The patent transitions from simple translation motion models to affine motion models that incorporate rotation, zoom, and perspective parameters. By changing the motion model parameters from basic translation vectors to affine transformation parameters (including rotation angles, zoom factors, and perspective coefficients), the system achieves higher prediction accuracy for complex motions
2Manufacturing precision
If affine motion modes are used for all blocks, then motion prediction accuracy improves, but device complexity and computational requirements increase
Solution Approach 1:
The patent implements dynamic selection of motion modes, where the encoder adaptsively chooses between affine and non-affine modes for different blocks based on motion complexity. This dynamic approach allows the system to use computationally intensive affine models only where necessary (improving accuracy) while using simpler models elsewhere (maintaining coding speed), resolving the contradiction between precision and productivity
Solution Approach 2:
Different motion models are applied to different regions (blocks) of the video frame based on local motion characteristics. Blocks with complex motion (zoom, rotation, perspective) use affine models for high accuracy, while blocks with simple translation use non-affine models for faster processing. This local differentiation optimizes both prediction accuracy and coding efficiency
3Productivity
If motion vectors are updated and used across subsequent blocks, then compression efficiency improves, but device complexity increases due to additional processing requirements
Solution Approach 1:
The patent performs preliminary motion compensation and updates motion vectors for subsequent blocks before final encoding. By pre-computing and storing updated motion information in a candidate list, the system reduces the computational burden during the main encoding process, improving compression efficiency while managing processing complexity through advance preparation
Solution Approach 2:
The system uses previously decoded motion information from neighboring blocks to generate motion vector predictions for current blocks. This self-service approach, where decoded data is reused for prediction, improves compression efficiency by reducing redundant information transmission while the complexity is managed through efficient data reuse rather than new computations
Data Source
AI summary
A method, a video processing apparatus and computer-readable program medium for video process is disclosed. The method of video coding including: using, during a conversion between a current block of video and a bitstream representation of the current block, an affine motion mode for the current block; updating motion information of the current block after motion compensation so as to obtain updated motion information of the current block; and using the updated motion information during conversion of subsequent video blocks and their respective bitstream representations.