Motion Candidate Lists for Geometry-Partitioned Video Blocks with Pruning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently handling motion vectors under geometry partitioning, leading to increased bandwidth demand and computational complexity.
Innovation Solution
The proposed techniques involve inserting motion candidates into a motion candidate list using specific derivation rules, including generating uni-prediction motion candidates, scanning regular motion candidates, and performing pruning steps to optimize motion vector prediction and encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion candidates are inserted using traditional derivation rules for geometry partitioned blocks, then motion vector prediction accuracy may be maintained, but computational complexity and bandwidth demand increase
Solution Approach 1:
The patent divides the current video block into multiple non-square/non-rectangular geometry partitioned regions, with each region having its own motion candidate list. This segmentation allows motion compensation to be performed independently for each partition, improving prediction accuracy for complex motion patterns while keeping the computational load distributed across partitions rather than requiring exhaustive search over the entire block
Solution Approach 2:
The patent applies different motion candidate derivation rules and pruning strategies to different geometry partitioned regions based on their local characteristics. Each partition can use customized motion search parameters and candidate selection criteria, allowing the system to adapt to local motion patterns and reduce overall computational complexity by focusing resources where they are most needed
2Ease of manufacture
If traditional motion candidate lists are used for geometry partitioned blocks, then encoding simplicity is maintained, but bandwidth demand and computational complexity increase
Solution Approach 1:
The patent extracts and removes redundant motion candidates from the candidate lists through pruning operations. By identifying and eliminating duplicate or highly correlated motion vectors before transmission, the system reduces the bandwidth required to transmit motion information while maintaining prediction accuracy, thus addressing the bandwidth demand issue without complicating the encoding process
Solution Approach 2:
The patent modifies the parameters of motion candidate lists for geometry partitioned blocks by applying specific derivation rules that generate fewer but more relevant candidates. By changing the candidate generation parameters and applying pruning thresholds, the system reduces the amount of motion data that needs to be transmitted and processed, thereby reducing bandwidth demand while keeping encoding straightforward
3Measurement precision
If comprehensive motion candidate lists are generated for all video blocks, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary pruning of motion candidate lists during the candidate generation phase, removing obviously redundant candidates before the actual motion estimation and encoding processes. By conducting this preliminary filtering action, the system reduces the number of candidates that need to be fully processed, thereby improving encoding speed without significantly compromising prediction accuracy
Solution Approach 2:
The patent applies partial pruning strategies where not all motion candidates are fully processed or transmitted. By selectively processing only the most promising candidates based on preliminary criteria, the system achieves sufficient prediction accuracy for most cases while significantly reducing computational complexity and improving encoding throughput
Data Source
AI summary
A video processing method including inserting motion candidates in a first motion candidate list of a video region of a first video block using a list derivation rule, where the video region has a geometry shape that is a non-square and non-rectangular shape, and the first video block is coded with a geometry partition mode; and performing a conversion between the first video block and a bitstream of the first video block using the first motion candidate list.