Motion Vector Refinement Scaling for Multi-Reference Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving accurate motion vector refinement, particularly when reference pictures are at varying distances from the current frame, leading to suboptimal coding accuracy and efficiency, especially in streaming contexts where subsequent frames are not available.
Innovation Solution
The implementation of decoder-side motion vector refinement (DMVR) techniques that utilize reference pictures at different distances and apply scaling factors to motion vectors, allowing for improved coding accuracy by using multiple reference pictures from the same side of the current frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If decoder-side motion vector refinement is performed using reference pictures at varying distances, then coding accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies scaling factors to motion vectors based on the temporal distance of reference pictures. By changing the parameter of motion vector magnitude according to reference picture distance, the system achieves accurate motion representation across different temporal levels while maintaining computational efficiency through a systematic parameter adjustment approach.
Solution Approach 2:
The patent divides the motion vector refinement process into distinct stages: initial motion vector derivation, scaling factor calculation based on reference picture distances, and refined motion vector computation. This segmentation allows each stage to be optimized independently, improving overall coding accuracy while managing computational complexity through structured processing.
2Measurement precision
If multiple reference pictures from the same side are used for DMVR, then coding accuracy improves and streaming context compatibility is achieved, but the process becomes more complex
Solution Approach 1:
The patent creates a universal DMVR framework that works across different coding contexts including streaming scenarios. By designing the scaling-based refinement process to handle various reference picture configurations (same-side and different-side references), the system achieves multi-functional compatibility while maintaining improved coding accuracy through consistent scaling principles.
3Measurement precision
If reference pictures at different temporal levels are used, then coding accuracy is enhanced, but temporal consistency becomes more difficult to maintain
Solution Approach 1:
The patent maintains temporal consistency across different temporal levels by dynamically adjusting the scaling factor parameter based on reference picture distances. This parameter change approach ensures that motion vectors remain temporally coherent even when combining references from multiple temporal levels, preserving stability while enhancing coding accuracy.
Data Source
AI summary
An example method of video coding includes receiving a video bitstream comprising a plurality of blocks and deriving a set of motion vectors for a current block of the plurality of blocks, the current block being within a current frame. The method also includes deriving a set of refined motion vectors for the current block using two or more reference pictures, the two or more reference pictures having different respective distances to the current frame. The method further includes reconstructing the current block based on the derived set of refined motion vectors.