Inter Prediction MMVD Candidate Checks for Motion Vector Overlap
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Merge mode with Motion Vector Difference (MMVD) technology in the Versatile Video Coding (VVC) standard often results in poor coding/decoding efficiency due to overlapping and close initial motion vectors, which reduces the effectiveness of inter prediction.
Innovation Solution
Implement a candidate check-based MMVD technique that checks for overlapping motion vectors by using a threshold value to determine if the distance between initial motion vectors is too close, and if so, reselects a new motion vector to avoid overlapping, thereby improving coding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the first two pieces of candidate motion information in a Merge candidate list are taken as MMVD candidate motion information, then the inter prediction process is simplified, but overlapping and close initial motion vectors reduce coding efficiency
Solution Approach 1:
The patent performs a preliminary check on the distance between initial motion vectors before using them for MMVD expansion. By calculating the distance in advance and comparing it with a threshold, the method prevents selecting overlapping motion vectors that would waste computational resources, thus improving coding efficiency without adding complex processing during the main prediction stage
Solution Approach 2:
The patent implements a feedback mechanism where the distance between motion vectors is evaluated and used to control the selection process. When the distance is below the threshold, the system adjusts the candidate selection to avoid overlapping vectors, creating a closed-loop control that optimizes the balance between simplicity and efficiency
2Adaptability or versatility
If motion vectors are expanded from multiple starting points, then more prediction candidates are generated, but overlapping vectors increase computational redundancy
Solution Approach 1:
Before expanding motion vectors from multiple starting points, the patent performs a preliminary distance check to ensure the starting points are sufficiently separated. This prevents the generation of overlapping expanded vectors, maintaining prediction diversity while avoiding computational redundancy in the expansion process
Solution Approach 2:
The patent applies different quality control to different stages of motion vector processing. The initial motion vectors undergo strict distance validation, while the expanded vectors benefit from the diversity of multiple starting points. This localized quality control ensures both diversity and efficiency without uniform overhead
Data Source
AI summary
A method and apparatus for inter prediction. According to the position of a current block in a video picture, first motion information and second candidate motion information is determined, the motion information at least includes a motion vector; it is determined whether the distance between the motion vector of the first motion information and the motion vector of the second candidate motion information is greater than a first threshold; if the determination result is that the distance is less than or equal to the first threshold, the second candidate motion information is updated to obtain second motion information; if the determination result is that the distance is greater than the first threshold, the second candidate motion information is taken as the second motion information; and inter prediction is performed on the current block by using the first motion information and the second motion information.