Motion Vector Prediction Candidate List Construction for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion vector prediction methods in moving picture coding systems increase memory access and processing load due to the use of different neighboring blocks for reference, leading to inefficient coding and increased decoding complexity.
Innovation Solution
A picture coding device and method that constructs candidate lists of reference block candidates in specific orders to reduce the number of reference blocks and optimize the selection of motion vector predictors, using both spatially neighboring blocks and blocks from different time frames to minimize memory access and processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If different neighboring blocks are used for motion vector prediction candidates, then the code size is reduced through better prediction accuracy, but the number of memory accesses and processing load increase
Solution Approach 1:
The patent segments the motion vector prediction process into two distinct modes: merge mode (using identical neighboring blocks) and AMVP mode (using different neighboring blocks). This segmentation allows the system to choose the appropriate mode based on coding conditions, thereby reducing the overall processing load while maintaining code size efficiency.
Solution Approach 2:
The patent dynamically switches between merge mode and AMVP mode based on the availability and suitability of neighboring blocks. The prediction mode is not fixed but adapts to the specific coding situation, optimizing the balance between code size reduction and processing complexity in real-time.
2Measurement precision
If multiple neighboring blocks are referenced for motion vector prediction, then prediction accuracy improves, but memory access operations increase
Solution Approach 1:
The patent extracts and uses only the necessary neighboring blocks for prediction based on the specific mode. In merge mode, only identical neighboring blocks are extracted and used, while in AMVP mode, different neighboring blocks are extracted. This selective extraction reduces the number of memory access operations compared to always using multiple different neighboring blocks.
Solution Approach 2:
The patent applies different prediction strategies to different local situations. When identical neighboring blocks are available and suitable, merge mode is applied locally. When different neighboring blocks provide better prediction, AMVP mode is applied locally. This local quality approach optimizes memory access based on the specific spatial and temporal context.
3Ease of manufacture
If motion information is coded in units of target blocks without reuse, then coding simplicity is maintained, but coding efficiency decreases due to duplicate information
Solution Approach 1:
The patent merges the motion information of the current block with that of identical neighboring blocks in merge mode. Instead of coding motion information independently for each block, the system merges with neighboring blocks when their motion information is identical or similar, thereby improving coding efficiency without significantly increasing coding complexity.
Solution Approach 2:
The patent makes motion information universal by allowing it to serve multiple blocks. When a neighboring block's motion information is used for the current block (in merge mode), the same motion information serves both blocks, eliminating redundancy and improving overall coding efficiency while maintaining relatively simple coding structures.
Data Source
Figure 1A~1B
Figure 2
Figure 3A~3B
AI summary
In a first inter-image prediction mode in which information on a motion vector is used, a first mode detection unit (106) constructs a first candidate list from a plurality of reference block candidates based on a first predetermined order, assigns indices for designating reference block candidates added in the first candidate list, and outputs the indices of the plurality of reference candidate blocks. In a second inter-image prediction mode in which is used a motion vector difference between a motion vector predictor based on information on a motion vector, and a motion vector of a coding target block, a second mode detection unit (103) constructs a second candidate list from a plurality of reference block candidates based on a second predetermined order, assigns indices for designating reference block candidates added in the second candidate list, and outputs the indices of the plurality of reference block candidates and motion vector differences.