Decoder Side Motion Vector Refinement Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently managing motion vectors, particularly in enabling or disabling decoder-side motion vector refinement (DMVR) and Bi-directional Optical flow (BIO) processes, which can lead to inaccurate motion information and reduced coding performance, especially when dealing with screen videos that often have integer motion vectors.
Innovation Solution
The proposed solution involves indicating the use of decoder motion information through flags in the bitstream representation, allowing for the enabling or disabling of DMVR and BIO processes based on initial motion vectors and reference pictures, thereby optimizing the video processing methods for specific conditions such as integer motion vectors or sub-pixel precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If decoder-side motion vector refinement (DMVR) is applied to all video blocks, then motion information accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent changes the parameter of motion vector precision by detecting whether the initial motion vector has sub-pixel precision. When sub-pixel precision is detected, the patent skips DMVR checks and uses the initial motion vector directly, thereby avoiding unnecessary computational complexity while maintaining motion information accuracy for sub-pixel cases.
2Measurement precision
If DMVR checks are performed for all motion vectors, then motion information accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs a preliminary check to detect the precision of the initial motion vector before conducting DMVR. This preliminary action allows the system to skip unnecessary DMVR checks when the initial motion vector already has sub-pixel precision, thereby reducing processing time while ensuring accuracy is maintained when needed.
Solution Approach 2:
The patent implements skipping of DMVR checks under specific conditions - when the initial motion vector has sub-pixel precision. This skipping mechanism eliminates redundant processing steps and reduces overall processing time without compromising motion information accuracy, as sub-pixel precision vectors already provide sufficient detail.
3Reliability
If BIO process is applied to all blocks, then coding performance is improved, but device complexity increases
Solution Approach 1:
The patent changes the parameter of motion vector precision detection to inform the BIO process decision. When sub-pixel precision is detected, the patent skips BIO checks and uses the initial motion vector directly, thereby reducing device complexity while maintaining coding performance for sub-pixel cases where BIO would not provide additional benefit.
4Productivity
If motion vector refinement processes are applied universally, then compression efficiency is improved, but loss of time increases
Solution Approach 1:
The patent performs a preliminary detection of initial motion vector precision before applying refinement processes. This preliminary action enables the system to skip unnecessary DMVR and BIO operations when sub-pixel precision is already present, thereby reducing encoding time while maintaining compression efficiency for cases where refinement would be beneficial.
Solution Approach 2:
The patent implements skipping of time-consuming refinement processes (DMVR and BIO) when the initial motion vector already has sub-pixel precision. This skipping mechanism eliminates redundant encoding operations, significantly reducing encoding time while preserving compression efficiency, as sub-pixel precision vectors already provide the necessary detail for effective compression.
Data Source
AI summary
Motion Vector Management for Decoder Side Motion Vector Refinement is disclosed. A method of video processing including: determining, for a conversion between a first block of video and a bitstream representation of the first block of video, whether and/or how to apply decoder-side motion vector refinement (DMVR) based on a signaled information; and performing the conversion based on the determination.