Geometric Partitioning for Lower-Complexity Video Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in achieving better compression ratios and lower complexity, particularly in handling motion vectors, with existing methods like triangle partition modes being computationally intensive and inefficient.
Innovation Solution
The introduction of geometric partitioning modes for video blocks, which utilize a set of angles and distances within threshold values, and rules based on block dimensions and coding characteristics to determine the applicability of geometric partitioning, allowing for more efficient video processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional motion vector handling methods are used, then video coding can be performed, but computational complexity is high and coding efficiency is insufficient
Solution Approach 1:
The current video block is divided into multiple prediction sub-regions using geometric partitioning modes (triangular, trapezoidal, quadrangular partitions). Each sub-region can have independent motion vectors, allowing for more precise motion representation while maintaining manageable computational complexity through regional decomposition
Solution Approach 2:
Different geometric partitioning modes are applied to different video blocks based on their specific characteristics (block size, motion complexity). The selection of partition types (triangular, trapezoidal, quadrangular) and the number of partitions are adapted locally to each block's requirements, optimizing coding efficiency without uniformly increasing complexity across all blocks
2Manufacturing precision
If geometric partitioning modes are applied to all video blocks, then coding precision is improved, but computational complexity increases
Solution Approach 1:
The geometric partitioning mode is dynamically selected for each video block based on its characteristics. The partitioning structure (number of regions, shape types) adapts to the local content requirements, enabling high prediction precision where needed while avoiding unnecessary complexity in simpler regions
Solution Approach 2:
Multiple geometric partitioning parameters are optimized including the number of partitions, partition shapes (triangular, trapezoidal, quadrangular), and motion vector precision. These parameters are adjusted based on block size and motion characteristics to achieve optimal balance between precision and complexity
Data Source
AI summary
A method of video processing is provided to comprise: performing a conversion between a current video block of a video and a bitstream representation of the video, wherein, during the conversion, a use of a geometric partitioning mode is allowed for the current video block, and wherein parameters of the geometric partitioning mode are computed using a set of angles including a first number of angles that is less than a first threshold value and/or a set of distances including a second number of distances that is less than a second threshold value.