Affine Motion Vector Coding for Complex Video Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques struggle to efficiently encode and decode video data with complex motion patterns, particularly in affine motion models, leading to suboptimal coding efficiency and image quality.
Innovation Solution
Implementing techniques for coding affine motion and affine motion models, including flag checking and advanced motion vector prediction (AMVP)-merge mode, to determine equal motion vector differences and generate affine candidate lists, thereby improving coding efficiency and image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional translational inter prediction methods are used, then device complexity is reduced, but coding efficiency and image quality deteriorate in affine motion scenarios
Solution Approach 1:
The motion model is segmented into multiple components: translational motion (first motion vector) and affine motion (second motion vector). The method divides the motion compensation process into stages, first applying simple translation and then applying affine transformation only where needed, thereby improving coding efficiency for complex motion while keeping the base complexity low
Solution Approach 2:
The motion model dynamically adapts between translational and affine modes based on motion complexity. The system transitions from a simple translational model to a more complex affine model when motion patterns require it, allowing the device complexity to scale dynamically with the actual motion characteristics of the video content
2Measurement precision
If affine motion models are implemented, then prediction accuracy improves, but bit requirements increase
Solution Approach 1:
The method applies affine motion compensation partially - only to the extent necessary for accurate prediction. By using a two-stage approach where translational motion is handled separately and affine transformation is applied as a补充, the system achieves high prediction accuracy while avoiding the full bit cost of pure affine models
Solution Approach 2:
The motion model parameters are changed dynamically based on motion characteristics. The system switches between using one motion vector (translational) and two motion vectors (affine), changing the parameter set size according to the actual motion complexity, thereby optimizing the balance between prediction accuracy and bit requirements
Data Source
AI summary
A video decoder may receive a block of video data to be decoded using a 6-parameter affine advanced motion vector predictor (AMVP) mode. The video decoder may decode a first syntax element indicating a first motion vector difference (MVD) for a first control point motion vector (CPMV) for the block, and also decode a flag that indicates if a second MVD for a second CPMV for the block and a third MVD for a third CPMV for the block are equal to the first MVD. The video decoder may further determine the second MVD and the third MVD based on the flag, and decode the block of video data using the first MVD, the second MVD, and the third MVD to generate a decoded block.