Regional Motion Prediction for Parallel Video Block Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding frameworks face strong dependencies between image blocks due to the use of history-based motion vector prediction, which hinders parallel processing and independent encoding/decoding of image blocks.
Innovation Solution
Utilize the motion information of encoded or decoded image blocks in a separate region to establish a historical motion information candidate list, allowing each image block in a region to be encoded or decoded independently, thereby reducing dependencies and enabling parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If history-based motion vector prediction is used to improve prediction accuracy, then encoding precision is improved, but dependency between image blocks increases
Solution Approach 1:
The patent divides the video data into independent processing units (tiles or slices) that can be encoded and decoded independently. By segmenting the video stream into separate regions with independent HMVP lists, the system maintains prediction accuracy within each segment while eliminating dependencies between segments, enabling parallel processing of different regions.
Solution Approach 2:
The patent applies different HMVP list management strategies to different regions of the video data. Each tile or slice maintains its own local HMVP list with region-specific motion information, allowing optimized prediction accuracy for local characteristics while preventing propagation of dependencies across the entire video stream.
2Manufacturing precision
If history-based motion vector prediction updates the HMVP list continuously, then prediction accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments the HMVP list management into independent regional lists rather than maintaining a single global list. Each regional HMVP list processes only local motion information, reducing the complexity of list management and updates while maintaining prediction accuracy within each region. This segmentation prevents the need to track and update dependencies across the entire video stream.
Solution Approach 2:
The patent pre-initializes separate HMVP lists for each tile or slice before encoding begins. By preparing these independent lists in advance with appropriate initial values, the system avoids complex runtime decisions about list management and enables straightforward parallel processing of multiple regions without coordination overhead.
3Manufacturing precision
If motion information from previously encoded blocks is used to update HMVP list, then encoding precision is improved, but parallel processing capability deteriorates
Solution Approach 1:
The patent divides the video stream into independent tiles or slices that can be processed in parallel. Each segmented region maintains its own HMVP list populated from motion information within that region only, eliminating cross-region dependencies and enabling simultaneous encoding of multiple regions without synchronization requirements.
Solution Approach 2:
The patent pre-populates HMVP lists for each parallel processing region with initial motion information before parallel encoding begins. This preliminary preparation ensures that each processing thread has the necessary prediction data independent of other threads, enabling true parallel execution while maintaining encoding precision through pre-established prediction capabilities.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A video processing method, a video processing device, and a computer storage media. The method includes: obtaining a historical motion information candidate list; and encoding or decoding each first image block of a plurality of first image blocks according to the historical motion information candidate list. The historical motion information candidate list is used for encoding or decoding each first image block of the plurality of first image blocks in a first region of a current frame, and the historical motion information candidate list is obtained based on motion information of second image blocks. The second image blocks are encoded or decoded image blocks other than image blocks in the first region.