Affine Motion Candidate Refinement for Complex Video Block Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in improving coding efficiency, particularly in handling complex motions such as zoom, rotation, and perspective motions, which are not adequately addressed by current affine motion models.
Innovation Solution
The proposed method involves determining an affine candidate list comprising refined and non-refined versions for video blocks, utilizing techniques like non-adjacent spatial neighbors, history-parameter-based models, and regression-based methods to enhance the construction of affine motion candidates, thereby improving the affine motion prediction in video coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a traditional affine motion model is used for video coding, then the implementation is simple, but the coding efficiency is insufficient especially for complex motions like zoom, rotation, and perspective motions
Solution Approach 1:
The patent segments the motion modeling process into multiple components: a traditional affine motion model for general motion and additional perspective motion models specifically for complex motions like zoom, rotation, and perspective transformations. This segmentation allows each model to specialize in specific motion types, improving overall coding efficiency without requiring the entire system to handle all complexity uniformly.
Solution Approach 2:
The patent introduces dynamic adaptation by selecting between different motion models (affine vs. perspective) based on the actual motion characteristics of the video block. The system dynamically determines which model to apply by analyzing motion vectors and comparing them against thresholds, allowing the motion compensation to adapt to varying motion complexities in real-time.
2Productivity
If more complex motion models are introduced to handle zoom, rotation, and perspective motions, then the coding efficiency improves, but the computational complexity increases
Solution Approach 1:
The patent applies partial action by only using complex perspective motion models when necessary, rather than always using the most complex model available. The system first attempts simpler affine motion modeling and only resorts to more computationally intensive perspective models when the motion characteristics require them, thus avoiding unnecessary computational complexity while maintaining compression efficiency for complex motions.
Solution Approach 2:
The patent changes parameters dynamically by adjusting the motion model selection based on motion vector magnitudes and other characteristics. When motion vectors indicate simple translation, the system uses simpler affine models with fewer parameters. When motion vectors indicate complex motions like zoom or rotation, the system switches to perspective models with additional parameters, optimizing the balance between compression performance and computational cost.
3Measurement precision
If a refined affine candidate list with multiple versions is constructed, then the motion prediction accuracy improves, but the processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-constructing multiple affine candidate lists with different versions (refined and non-refined) during the encoding process. These candidate lists are prepared in advance and stored for later use during decoding, allowing the system to avoid time-consuming real-time refinements while still benefiting from accurate motion prediction through the pre-computed refined candidates.
Solution Approach 2:
The patent introduces dynamic selection during decoding by choosing between refined and non-refined affine candidates based on actual decoding requirements and available reference pictures. The system dynamically determines which candidate list version to use based on the decoding context, balancing prediction accuracy needs with processing time constraints in real-time.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. In the method, for a conversion between a current video block of a video and a bitstream of the video, an affine candidate list of the current video block is determined. The affine candidate list comprises a refined version of an affine candidate and a non-refined version of the affine candidate. The conversion is performed based on the affine candidate list.