HMVP Candidate Pruning for Merge and AMVP Motion Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in efficiently encoding motion information, particularly in merge modes, leading to suboptimal compression ratios and increased bandwidth usage.
Innovation Solution
Implementing a method for video processing that involves maintaining a set of tables with motion candidates, applying a pruning operation to update a candidate list, and performing conversions between video blocks and bitstream representations using the constructed candidate list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion candidates are added to the merge/AMVP list without pruning, then the candidate list becomes more comprehensive for motion prediction, but the list size increases leading to higher complexity and reduced encoding efficiency
Solution Approach 1:
The patent changes the parameter of candidate list management by introducing dynamic pruning operations that adjust the list based on motion information characteristics. Tables store motion candidates with associated motion information, and pruning operations selectively remove redundant candidates based on motion vector differences and reference picture indices, optimizing the balance between comprehensiveness and complexity.
Solution Approach 2:
The patent implements discarding of redundant motion candidates through pruning operations while recovering useful motion information by storing it in tables. The pruning operation removes duplicate or less useful candidates from the merge/AMVP list, while the table structure preserves motion candidates for potential future use, enabling efficient list management without permanent loss of valuable motion data.
2Adaptability or versatility
If redundant motion candidates are included in the candidate list, then more motion options are available for prediction, but bandwidth is wasted transmitting unnecessary motion information
Solution Approach 1:
The patent changes the parameter of candidate list composition by dynamically pruning redundant motion candidates based on motion information comparison. The pruning operation evaluates motion vectors and reference picture indices to identify and remove redundant candidates, reducing the number of candidates that need to be transmitted while maintaining adaptability through the table-based storage system that preserves motion information for potential recovery.
Solution Approach 2:
The patent extracts and removes redundant motion candidates from the candidate list through pruning operations. By comparing motion information between candidates and extracting only the essential unique motion data, the system reduces bandwidth consumption while maintaining the necessary motion prediction options. The extraction process separates useful motion information from redundant duplicates.
3Measurement precision
If a large number of motion candidates are stored in tables, then the motion prediction becomes more accurate, but the memory requirements and processing complexity increase
Solution Approach 1:
The patent changes the parameter of table storage by implementing pruning operations that dynamically adjust the number of stored motion candidates based on redundancy analysis. The system maintains high motion vector precision by preserving unique motion information in tables while removing redundant entries, optimizing the balance between precision and storage requirements through conditional pruning based on motion vector and reference picture comparisons.
Solution Approach 2:
The patent discards redundant motion candidate storage in tables while recovering essential motion information. The pruning operation identifies and removes duplicate motion candidates from table storage, reducing memory requirements. Simultaneously, the system recovers and preserves unique motion information in the tables, ensuring that only necessary data is stored with optimized memory utilization.
Data Source
AI summary
A method of video processing is provided to include: maintaining a set of tables, wherein each table includes motion candidates and each motion candidate is associated with corresponding motion information; updating a motion candidate list based on motion candidates in one or more tables using a pruning operation on the motion candidates; and performing a conversion between a first video block and a bitstream representation of a video including the first video block using the constructed motion candidate list.