Decoder-Side Motion Refinement for Low-Latency Bi-Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like HEVC face limitations in achieving superior coding efficiency, and the emerging Versatile Video Coding (VVC) standard requires improvements in motion vector refinement to enhance compression efficiency.
Innovation Solution
Implement decoder-side motion vector refinement (DMVR) techniques to refine non-refined motion vectors, using spatial and temporal prediction methods, and optical flow models to generate refined motion vectors for improved bi-prediction in video coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If decoder-side motion vector refinement (DMVR) is implemented to improve compression efficiency, then coding efficiency is improved, but coding latency increases
Solution Approach 1:
The patent divides the picture into multiple segments and processes motion vector refinement segment by segment. This segmentation allows the decoder to process only the current segment using refined motion vectors from previous segments, rather than waiting for the entire picture to be processed, thereby reducing coding latency while maintaining compression efficiency gains from DMVR.
Solution Approach 2:
The patent performs motion vector refinement preliminarily for current blocks using motion vectors from previously decoded blocks. By preparing refined motion vectors in advance for blocks that can be predicted from already-decoded content, the system reduces the waiting time for motion compensation without sacrificing the quality improvements from full DMVR processing.
2Measurement precision
If motion vector refinement is performed using spatial and temporal prediction methods, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies motion vector refinement partially by selectively applying it only to certain blocks or regions where it provides the most benefit, rather than uniformly processing all blocks. This partial application maintains prediction accuracy for critical areas while reducing overall computational complexity.
Solution Approach 2:
The patent applies different motion vector refinement strategies to different regions of the picture based on their characteristics. Areas with higher motion complexity receive more aggressive refinement, while stable regions use simpler prediction methods, optimizing the balance between prediction accuracy and computational complexity locally.
Data Source
AI summary
Embodiments of video coding systems and methods are described for reducing coding latency introduced by decoder-side motion vector refinement (DMVR). In one example, two non-refined motion vectors are identified for coding of a first block of samples (e.g. a first coding unit) using bi-prediction. One or both of the non-refined motion vectors are used to predict motion information for a second block of samples (e.g. a second coding unit). The two non-refined motion vectors are refined using DMVR, and the refined motion vectors are used to generate a prediction signal of the first block of samples. Such embodiments allow the second block of samples to be coded substantially in parallel with the first block without waiting for completion of DMVR on the first block. In additional embodiments, optical-flow-based techniques are described for motion vector refinement.