Multi-Layer Motion Vector Refinement for Efficient Image Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution, high-quality images and videos, particularly in virtual reality and augmented reality, necessitates a more efficient image/video compression technique to reduce transmission and storage costs while maintaining quality.
Innovation Solution
Implementing a multi-layer Decoder-side Motion Vector Refinement (MDMVR) method for improved image coding efficiency, which includes determining the use of MDMVR for current blocks, deriving refined motion vectors, and generating reconstructed samples based on these vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution and high quality video/image formats are used, then image quality and resolution are improved, but transmission and storage costs increase
Solution Approach 1:
The patent extracts and transmits only the essential information for motion representation by using motion vector refinement. Instead of transmitting all pixel data for high-resolution video, the system extracts motion vectors and refines them at the decoder side, significantly reducing the data amount that needs to be transmitted while maintaining high image quality.
Solution Approach 2:
The patent performs preliminary motion estimation at the encoder side to generate initial motion vectors, and then performs refinement at the decoder side. This preliminary action allows the encoder to prepare basic motion information without performing computationally intensive refinement operations, reducing transmission data while enabling high-quality reconstruction.
2Measurement precision
If motion vector refinement is applied to improve prediction accuracy, then inter-coding efficiency is improved, but processing complexity increases
Solution Approach 1:
The patent segments the motion vector processing into two distinct stages: initial motion estimation at the encoder and refinement at the decoder. This segmentation allows the complex refinement operations to be performed only at the decoder side using already-transmitted motion vector data, improving accuracy without proportionally increasing overall system complexity.
Solution Approach 2:
The decoder performs motion vector refinement using the transmitted motion vector data and local picture information. This self-service approach allows the decoder to improve motion vector accuracy independently without requiring additional complex processing at the encoder, effectively distributing the computational burden.
3Productivity
If multi-layer MDMVR is used to enhance inter-coding performance, then compression efficiency is improved, but processing complexity increases
Solution Approach 1:
The patent implements dynamic multi-layer MDMVR processing where the refinement process adapts based on the specific block characteristics and available data. The system dynamically determines the appropriate level of refinement and processing depth for different regions, improving compression efficiency while avoiding unnecessary complex processing in areas where it is not needed.
Solution Approach 2:
The patent applies different levels of motion vector refinement to different blocks or regions based on their specific characteristics. High-refinement processing is applied only where it provides the most benefit, while simpler processing is used elsewhere, improving overall compression efficiency without uniformly increasing processing complexity across the entire image.
Data Source
AI summary
An image decoding method performed by a decoding device is provided. The method comprises the steps of: determining whether to use a multi-layer decoder-side motion vector refinement (MDMVR) for a current block; deriving a refined motion vector for the current block on the basis of using the MDMVR for the current block; deriving prediction samples for the current block on the basis of the refined motion vector; and generating reconstruction samples for the current block on the basis of the prediction samples.