Affine Model Bilateral Matching for Complex Video Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently utilizing intra prediction and motion compensation techniques to reduce redundancy in video data, particularly in handling complex motion patterns and directions, which affects compression efficiency and bitrate requirements.
Innovation Solution
The implementation of affine bilateral matching using affine models for video encoding/decoding, which involves determining first and second affine parameters for reference pictures and deriving control point motion vectors (CPMVs) to reconstruct blocks based on affine motion, utilizing 3- or 4-parameter models to enhance motion vector prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional motion compensation techniques are used, then computational complexity is reduced, but compression efficiency deteriorates when handling complex motion patterns
Solution Approach 1:
The current block is divided into multiple sub-blocks, and affine motion parameters are derived independently for each sub-block. This segmentation allows the system to handle complex motion patterns with higher precision while maintaining manageable computational complexity by processing smaller, independent regions rather than the entire block at once.
Solution Approach 2:
The patent transitions from traditional translational motion models to affine motion models that operate in an expanded parameter space. By introducing additional motion parameters (rotation, scaling, shearing) beyond simple translation, the system captures complex motion patterns more accurately, effectively adding dimensional complexity to the motion representation.
2Measurement precision
If affine motion models with multiple parameters are used, then motion vector prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary derivation of affine motion parameters from control points at the block boundaries before generating motion vectors for the entire block. This preliminary action establishes accurate motion models early in the process, enabling more precise motion vector prediction while reducing the computational burden of subsequent processing steps.
Solution Approach 2:
The system dynamically adjusts the number of affine motion parameters based on the complexity of the motion pattern detected in each block. For simple motion, fewer parameters are used, while for complex motion patterns, the full affine model with multiple parameters is applied. This adaptive parameter selection maintains high accuracy when needed while reducing complexity for simpler cases.
Data Source
AI summary
First affine parameters and second affine parameters for each of one or more affine models associated with a current block are determined. The first affine parameters of the respective affine model are associated with a first motion vector for a first reference picture of the current block. The second affine parameters of the respective affine model are associated with a second motion vector for a second reference picture of the current block. The first affine parameters and the second affine parameters of the respective affine model have opposite signs. Control point motion vectors (CPMVs) of the affine motion of the current block are determined by performing the affine bilateral matching based on the determined first affine parameters and the determined second affine parameters of the one or more affine models. The current block is reconstructed based on the CPMVs of the affine motion of the current block.