Geometric Partition Motion Compensation With Shift-Based Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques face high computational complexity due to the need for complex calculations in geometric partition modes, leading to inefficiencies in video encoding and decoding processes.
Innovation Solution
Implement techniques that reduce computational complexity by determining weighting values for geometric partition modes using power-of-2 numbers, eliminating the need for sine and cosine calculations, and applying shift operations instead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex sine and cosine calculations are used for geometric partition modes, then prediction precision is improved, but computational complexity increases
Solution Approach 1:
The patent changes the parameter representation from trigonometric functions (sine and cosine) to power-of-2 numbers. This parameter transformation allows the same geometric partitioning function to be achieved with simpler arithmetic operations, reducing computational complexity while maintaining prediction precision through equivalent mathematical relationships
Solution Approach 2:
The patent substitutes complex trigonometric calculations with simpler shift operations. By replacing the mechanical computation of sine and cosine with bitwise shift operations based on power-of-2 numbers, the system achieves the same geometric partitioning effect with significantly reduced computational burden
2Measurement precision
If detailed motion field storage is implemented, then motion compensation accuracy is improved, but memory requirements and processing complexity increase
Solution Approach 1:
The patent extracts only the essential motion information needed for compensation by representing motion fields using power-of-2 numbers. This extraction approach captures the critical motion characteristics while eliminating redundant data, reducing storage requirements without sacrificing motion compensation accuracy
Solution Approach 2:
The patent transforms the storage format of motion field data from traditional floating-point or fixed-point representations to power-of-2 numbers. This parameter change enables more efficient memory utilization and simplifies processing while preserving the accuracy needed for effective motion compensation
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
A video decoder can be configured to determine, for a block of video data encoded in a geometric partition mode, an angle for the block for the geometric partition mode; determine a separation line displacement relative to a center of the block for the geometric partition mode; partition the block into first and second partitions based on the angle and the separation line displacement; determine first predictive samples for the block using a motion vector for the first partition and second predictive samples for the block using a motion vector for the second partition; determine a power-of-2 number based on the angle for the block; determine weighting values based on the power-of-2 number; perform a blending operation on the first predictive samples and the second predictive samples based on the weighting values to determine a prediction block for the block.