History-Based Motion Vector Prediction Buffer for Efficient Inter Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high resolution and high quality images/video, such as 4K or 8K Ultra High Definition, along with the growth of virtual reality and augmented reality content, necessitates a highly efficient image/video compression technique to reduce transmission and storage costs while managing diverse image/video characteristics.
Innovation Solution
The implementation of a history-based motion vector prediction (HMVP) method and apparatus, including an HMVP buffer management system for efficient inter prediction, which updates and initializes the buffer based on motion information of previous blocks, supports parallel processing, and enhances coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution and high quality image/video data are transmitted, then image quality and detail are improved, but transmission cost and storage cost increase
Solution Approach 1:
The patent extracts and transmits only the residual information (difference between original and predicted blocks) rather than the complete image data. By removing the prediction component and transmitting only the residual, the data量 is significantly reduced while maintaining image quality, directly resolving the contradiction between quality and transmission cost
Solution Approach 2:
The patent introduces an intermediary prediction mechanism that generates predicted blocks from reference blocks and motion vectors. This intermediary step allows the system to represent image data more efficiently by transmitting only the residual information needed to reconstruct the image, reducing transmission costs while preserving quality
2Measurement precision
If history-based motion vector prediction buffer is updated with every block, then prediction accuracy is improved, but processing time and complexity increase
Solution Approach 1:
The patent applies different update strategies to different blocks based on their position and characteristics. Not all blocks trigger HMVP buffer updates - only specific blocks (e.g., those with certain motion characteristics or positions) are processed, allowing the system to maintain prediction accuracy where needed while reducing processing time elsewhere
Solution Approach 2:
The patent uses partial updates to the HMVP buffer rather than complete updates for every block. The buffer is updated with motion vector information from processed blocks, but the update process is optimized to avoid excessive processing, balancing prediction accuracy improvement with computational efficiency
3Reliability
If HMVP buffer is initialized at every CTU, then prediction reliability is improved, but initialization overhead increases
Solution Approach 1:
The patent segments the picture into CTUs (Coding Tree Units) and applies HMVP buffer initialization selectively at specific segmentation points (CTU boundaries) rather than continuously. This segmentation approach ensures prediction reliability is maintained at natural image structure boundaries while reducing the frequency of initialization operations
Solution Approach 2:
The patent performs HMVP buffer initialization in advance at predetermined CTU positions before processing begins. By pre-initializing the buffer at strategic locations rather than continuously, the system ensures prediction reliability is established when needed while minimizing initialization overhead through advance preparation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to an embodiment disclosed in the present document, a history-based motion vector prediction (HMVP) buffer for the current block can be derived on the basis of a history, and motion information of the current block can be derived on the basis of HMVP candidates included in the HMVP buffer, and thus inter prediction efficiency can be increased.