Symmetric MVD Motion Prediction With Decoder-Side Vector Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in encoding and decoding processes due to limitations in motion vector prediction, particularly with symmetric motion vector differences, which restrict the selection of non-symmetrical motion vectors, even if they offer slight advantages.
Innovation Solution
The implementation of decoder-side motion vector refinement (DMVR) in conjunction with advanced motion vector prediction (AMVP) allows for the formation of refined motion vectors, enabling the use of non-symmetrical motion vectors while skipping the signaling of certain reference indices and motion vector differences, thereby improving coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If symmetric motion vector difference mode is used to skip signaling of reference indices and motion vector differences, then bitstream complexity is reduced, but the ability to select non-symmetrical motion vectors is lost
Solution Approach 1:
The patent applies preliminary action by performing motion vector refinement at the decoder side before final motion vector determination. The decoder refines the base motion vectors using DMVR process, allowing non-symmetrical motion vectors to be selected while still benefiting from the simplified signaling of symmetric motion vector difference mode. This preliminary refinement step enables the system to overcome the limitation of symmetric MVD mode.
2Adaptability or versatility
If all motion vector information is signaled to allow non-symmetrical motion vectors, then motion vector selection flexibility is improved, but bitstream complexity increases
Solution Approach 1:
The patent introduces an intermediary approach by using symmetric motion vector difference mode as a base and adding a refinement step. Instead of signaling all motion vector information explicitly, the system uses the symmetric MVD as an intermediary representation and allows the decoder to refine it to achieve non-symmetrical motion vectors when needed. This mediator approach balances between compression efficiency and motion vector flexibility.
3Productivity
If symmetric motion vector difference mode is used, then encoding efficiency is improved by skipping signaling, but video quality may deteriorate due to restricted motion vector options
Solution Approach 1:
The patent applies feedback by implementing decoder-side motion vector refinement that uses the refined motion vectors to improve prediction accuracy. The refinement process provides feedback to adjust the motion vectors, allowing the system to maintain high encoding efficiency while improving video quality through more accurate motion compensation. This feedback mechanism ensures that the restricted symmetric MVD signaling does not permanently limit the quality.
Data Source
AI summary
A video coder is configured to form, in a symmetric motion vector difference mode, a List 0 (L0) base vector using a L0 Advanced Motion Vector Prediction (AMVP) candidate list and a List 1 (L1) base vector using a L1 AMVP candidate list; determine a refined L0 motion vector and a refined L1 motion vector by performing a decoder-side motion vector refinement process that refines the L0 base vector and the L1 base vector; and use the refined L0 motion vector and the refined L1 motion vector to determine a prediction block for a current block of a current picture of the video data.