HMVP Table Maintenance for Non-Rectangular Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently processing non-rectangular and non-square video blocks, particularly in geometry partitions, where existing methods struggle with motion vector prediction and coding efficiency.
Innovation Solution
The implementation of history-based motion vector prediction (HMVP) techniques that allow for the use of non-rectangular and non-square sub-portions within video blocks, enabling more efficient processing by deriving motion information from previously coded blocks and using current images as references, and incorporating HMVP candidates into merge lists for improved prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional rectangular and square video blocks are used for motion vector prediction, then the processing is simpler and more compatible with existing standards, but the coding efficiency is reduced when processing non-rectangular video content
Solution Approach 1:
The video block is divided into multiple sub-portions, where each sub-portion can be independently processed with appropriate motion prediction techniques. This segmentation allows non-rectangular regions to be handled separately, improving coding efficiency without requiring complete redesign of the processing system.
Solution Approach 2:
The patent introduces dynamic determination of sub-portion configurations based on the specific video content and coding conditions. The system adaptively selects whether to process blocks as traditional rectangular units or divide them into non-rectangular sub-portions, balancing coding efficiency gains against processing complexity in real-time.
2Measurement precision
If non-rectangular and non-square sub-portions are used within video blocks, then motion prediction accuracy is improved for complex video content, but the processing complexity and computational load increase
Solution Approach 1:
Different sub-portions of the video block are processed with different motion prediction strategies tailored to their local characteristics. Non-rectangular sub-portions requiring higher prediction accuracy receive more sophisticated processing, while simpler regions use standard methods, optimizing the balance between accuracy and complexity.
Solution Approach 2:
The patent modifies traditional motion prediction parameters by introducing sub-portion-specific motion vectors and reference image selections. This allows the system to achieve higher prediction accuracy for non-rectangular regions by using customized parameters rather than forcing all regions into a uniform rectangular framework.
3Reliability
If HMVP candidates are stored and maintained for non-rectangular blocks, then motion vector prediction quality is enhanced, but the memory requirements and table maintenance complexity increase
Solution Approach 1:
The HMVP table is segmented into multiple sub-portions corresponding to different regions of the video block. Each sub-portion maintains its own motion vector candidates, allowing the system to store only the necessary prediction data for each specific region rather than maintaining a single large table for the entire block.
Solution Approach 2:
The HMVP table structure is designed to serve multiple functions: it stores motion vectors for both traditional rectangular blocks and non-rectangular sub-portions, and it can be selectively updated based on the coding mode being used. This multi-functionality reduces the need for separate maintenance procedures for different block types.
4Adaptability or versatility
If current images are used as reference images for non-rectangular sub-portions, then the flexibility and accuracy of motion prediction is improved, but the computational resources required for processing increase
Solution Approach 1:
The system applies current image as reference only to the extent necessary for non-rectangular sub-portions where it provides significant benefit, rather than universally applying it to all blocks. This partial application approach achieves the flexibility and accuracy improvements where needed while limiting the additional computational energy consumption.
Data Source
AI summary
The disclosure relates to table maintenance for HMVP candidate storage. A method for processing video, including: determining, during a conversion between a first block of visual media data and a corresponding coded representation of the visual media data, the first block being coded with geometry partition mode; determining, based on at least one table storing history based motion vector prediction (HMVP) candidates which include motion information based on previously coded blocks, motion information of at least one sub-portion of the first block; performing the conversion of the first block using the determined motion information.