Affine Motion Vector Derivation for Sub-Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges with affine motion compensation, including increased memory requirements, precision loss in motion vector derivation, and inefficient construction of sub-block motion candidate lists, particularly in affine merge and inter modes.
Innovation Solution
The proposed solution involves optimizing affine motion compensation by reducing memory requirements, improving motion vector derivation precision, and adaptively managing sub-block motion candidate lists through techniques such as affine model inheritance, direct derivation of sub-block motion vectors from neighboring blocks, and adaptive ordering of candidate lists based on coded block information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If affine motion compensation is applied to improve motion prediction accuracy, then motion vector precision is improved, but memory requirements increase
Solution Approach 1:
The patent extracts and stores only the necessary control point motion vectors (CP-MVs) at specific locations rather than storing all motion vectors. This selective extraction reduces memory requirements while maintaining the precision needed for affine motion compensation by deriving other motion vectors from these stored CP-MVs through affine transformation equations.
Solution Approach 2:
The current block is divided into multiple sub-blocks, each with its own motion vector derived from control points. This segmentation allows the system to use fewer control points per sub-block while maintaining overall precision, thereby reducing total memory requirements compared to storing full motion vectors for every pixel or large block.
2Reliability
If control point motion vectors are derived and stored for predicting other blocks, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary derivation and storage of control point motion vectors at specific locations before actual motion compensation. These pre-computed CP-MVs are stored in a candidate list and reused for predicting multiple subsequent blocks, which improves prediction accuracy while reducing the computational complexity of real-time processing by avoiding redundant calculations.
3Productivity
If sub-block motion candidate lists are constructed using traditional methods, then motion information is obtained, but computational complexity increases
Solution Approach 1:
The patent applies partial action by deriving motion vectors only for necessary sub-blocks using a simplified approach. Instead of performing full affine transformations for all sub-blocks, the system uses direct derivation from neighboring block CP-MVs for certain sub-blocks, reducing computational complexity while maintaining sufficient coding efficiency for those regions.
Data Source
AI summary
A method of video processing includes: determining an affine mode type to be applied to a current block coded in an affine mode in a video frame, wherein the current block is split into a plurality of sub-blocks; and deriving motion information for each of the plurality of sub-blocks based on the affine mode type, at least one control point (CP) motion vector (MV) at a specific location being derived and stored for predicting other blocks to be processed subsequently.