SMVD Reference Indexing With Short-Term Motion Vector Signaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution, high-quality image/video data, particularly in immersive media formats like VR and AR, has led to higher transmission and storage costs due to the increased amount of information, necessitating a more efficient image/video compression technology.
Innovation Solution
The method and device improve image/video coding efficiency by optimizing inter prediction processes, including signaling information on motion vector differences, particularly L0 and L1 motion vector differences, and utilizing short-term reference pictures for deriving SMVD reference indices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution, high-quality image/video data is transmitted or stored, then image/video quality is improved, but transmission cost and storage cost increase
Solution Approach 1:
The patent applies parameter changes by optimizing motion vector difference signaling through bi-prediction with CU-level weight (BCW). The system changes the precision and representation parameters of motion vectors by using weighted prediction with coefficients alpha and beta, allowing efficient compression that maintains quality while reducing data size. This resolves the contradiction by changing how motion information is represented and transmitted.
Solution Approach 2:
The patent replaces traditional mechanical compression methods with advanced inter-prediction techniques involving motion vector optimization. By substituting conventional compression approaches with bi-prediction and motion vector difference signaling, the system achieves better compression efficiency without sacrificing quality.
2Productivity
If inter prediction with SMVD and MMVD is performed, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the prediction process into distinct components: L0 prediction using short-term reference pictures and L1 prediction using long-term reference pictures. The motion vector difference is segmented into components that can be independently optimized and signaled, reducing the overall complexity while maintaining efficiency.
Solution Approach 2:
The patent introduces dynamic adaptation through bi-prediction with CU-level weight, where prediction coefficients (alpha and beta) are adaptively determined based on block characteristics. This dynamic approach allows the system to optimize coding efficiency for each block while managing complexity through conditional application rather than fixed complex processing.
3Productivity
If motion vector difference signaling is optimized for bi-prediction, then inter prediction efficiency is improved, but signaling complexity increases
Solution Approach 1:
The patent applies local quality by providing separate optimization for L0 and L1 motion vector differences, with specific signaling for bi-prediction cases. The system tailors the signaling parameters locally for different prediction directions and block types, improving efficiency where needed while keeping signaling simple where standard procedures apply.
Data Source
AI summary
According to embodiments according to the present document, merge mode motion vector differences (MMVD) and symmetric motion vector differences (SMVD) may be performed on the basis of reference picture types and, in particular, short-term reference pictures may be used. Accordingly, prediction performance and coding efficiency in an inter prediction mode may be increased.