Geometric Partitioning Template Matching for VVC Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The implementation of adaptive blending in the Geometric Partitioning Mode (GPM) of the Versatile Video Coding (VVC) standard requires further improvements to enhance the accuracy and efficiency of motion prediction.
Innovation Solution
Extensions to template matching for geometric partitioning, including the application of template blending at a splitting line, reordering of partitioning modes by template matching cost, and bitstream signaling of blending area width, are introduced to improve the accuracy and efficiency of motion prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If adaptive blending is implemented in GPM, then motion prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The current block is divided into two partitions using a geometric partitioning line, with template matching performed separately on each partition. This segmentation allows the complex adaptive blending process to be broken down into manageable steps, improving motion prediction accuracy while controlling computational complexity through localized processing.
Solution Approach 2:
Template matching is performed in advance to determine motion vectors for each partition before the final blending step. This preliminary action prepares the necessary motion compensation data, allowing the adaptive blending to focus only on combining the pre-computed partitions, thereby improving accuracy without proportionally increasing overall computational complexity.
2Measurement precision
If template matching is performed on extended templates, then motion vector refinement accuracy is improved, but processing time increases
Solution Approach 1:
The template matching is performed locally on each partition using the geometric partitioning line as a reference. By focusing template matching on local regions rather than the entire block, the patent achieves refined motion vectors for each partition while reducing the overall processing time through localized computation.
Solution Approach 2:
Template matching is applied selectively to the partition boundaries and edges where geometric partitioning is most beneficial, rather than uniformly across the entire block. This partial application of template matching refines motion vectors where they are most needed while minimizing unnecessary processing time expenditure.
3Measurement precision
If blending area width is increased, then motion prediction accuracy at partition boundaries is improved, but computational load increases
Solution Approach 1:
The adaptive blending process applies different blending area widths to different regions based on local characteristics. Larger blending areas are applied only at partition boundaries where geometric partitioning creates discontinuities, while smaller or no blending is applied in uniform regions. This localized approach improves boundary prediction accuracy without uniformly increasing computational load across the entire block.
Data Source
AI summary
A VVC-standard encoder and a VVC-standard decoder are provided, configuring one or more processors of a computing system to perform template matching for geometric partitioning, including extensions of template size for template matching, application of blending at a splitting line, reordering of partitioning modes by template matching cost and blending area width, and bitstream signaling of blending area width.