HMVP Table Sizing for Non-Rectangular Video Block Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in efficiently handling non-rectangular and non-square video blocks, particularly in terms of motion vector prediction, leading to suboptimal coding efficiency and increased computational complexity.
Innovation Solution
The implementation of geometry partition modes, including history-based motion vector prediction (HMVP), which allows for the division of video blocks into non-rectangular and non-square sub-portions, and the use of HMVP candidates to enhance motion vector prediction, particularly in standards like HEVC and VVC, by improving coding efficiency and reducing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If geometry partition modes with HMVP are implemented for non-rectangular blocks, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The video block is divided into multiple geometry partitions (triangular, non-rectangular shapes) instead of traditional rectangular blocks. Each partition can be independently processed with its own motion vector prediction, allowing better adaptation to content characteristics while maintaining manageable complexity through systematic segmentation rules
Solution Approach 2:
HMVP (History-based Motion Vector Prediction) pre-calculates and stores motion vector candidates from previously processed blocks before processing the current block. This preliminary action provides ready-to-use prediction candidates, reducing the computational burden during actual encoding/decoding operations and improving coding efficiency without significant complexity increase
2Device complexity
If traditional rectangular block processing is used, then device complexity is lower, but coding efficiency for non-rectangular content is reduced
Solution Approach 1:
Different geometry partitions within the same video block are assigned different coding characteristics and motion vector prediction methods based on their specific shapes and positions. Triangular partitions use one prediction approach while non-rectangular partitions use another, optimizing coding efficiency for each local region without requiring complete system redesign
Solution Approach 2:
The system dynamically selects between different geometry partition modes (rectangular, triangular, non-rectangular) based on the content characteristics and block position. This dynamic adaptation allows the encoder to choose the most efficient partitioning strategy for each specific case, improving overall coding efficiency while maintaining manageable complexity through conditional logic
Data Source
AI summary
The disclosure relates to improvements for a history based motion vector prediction (HMVP) table. A method for processing video includes maintaining, during a conversion between a current block of visual media data and a corresponding coded representation of the visual media data, a table storing HMVP candidates which include motion information based on previously coded blocks. A conversion of the current block is performed at least based on the table. The table has a HMVP table size which depends on one or more motion candidate numbers numHMVPs added to one or more motion candidate lists. The HMVP table size is a maximum number of candidates stored in the table.