Image Decoding Prediction Block Combination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In moving image compression techniques, bi-directional inter prediction results in increased prediction error and deteriorated compression performance due to splitting of prediction units and additional motion information.
Innovation Solution
The method involves using multiple partitioning modes for inter prediction, applying different modes for each reference picture, and combining prediction blocks of varying shapes to reduce prediction error and additional information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If bi-directional inter prediction is performed by averaging prediction blocks of the same type, then motion information can be utilized from both reference pictures, but prediction error increases and compression performance deteriorates due to splitting of prediction units and increase of additional information amount
Solution Approach 1:
The current block is divided into multiple prediction units with different partitioning modes (first partitioning mode and second partitioning mode). Each prediction unit independently processes motion information from reference pictures, allowing flexible adaptation to different motion patterns in different regions without forcing uniform averaging across the entire block.
Solution Approach 2:
Different prediction units within the same current block can use different partitioning modes tailored to their local characteristics. The MPM list is configured according to the first partitioning mode, allowing each region to select the most appropriate partitioning mode for its specific motion patterns, thereby improving local prediction accuracy without compromising overall compression.
2Reliability
If prediction units are split to reduce prediction error, then prediction accuracy improves, but additional information amount including motion information increases
Solution Approach 1:
Multiple prediction units with different partitioning modes are combined to form the final prediction block. The MPM list mechanism allows the decoder to reconstruct the prediction block by merging predictions from different partitioning modes without requiring transmission of all individual partitioning mode data, thereby reducing the additional information amount while maintaining prediction accuracy.
Solution Approach 2:
The partitioning mode selection is dynamic and adaptive. The MPM list is configured according to the first partitioning mode, allowing the system to dynamically select the second partitioning mode based on local characteristics. This dynamic adaptation enables accurate prediction without committing to a fixed, information-heavy structure.
3Ease of manufacture
If a single partitioning mode is used for inter-picture prediction, then processing is simplified, but compression performance deteriorates due to inability to adapt to different motion patterns in different regions
Solution Approach 1:
The system supports multiple partitioning modes (first and second partitioning modes) within a unified prediction framework. The MPM list mechanism provides a universal interface that can accommodate different partitioning modes, allowing the same processing structure to handle diverse motion patterns across different regions of the current block.
Data Source
AI summary
An image decoding method may include: deriving motion information of a first prediction unit determined by a first partitioning mode of a current block and motion information of a second prediction unit determined by a second partitioning mode of the current block; generating a prediction block of the first prediction unit by using the motion information of the first prediction unit and generating a prediction block of the second prediction unit by using the motion information of the second prediction unit; and generating a prediction block of the current block in units of sub-prediction unit determined by combining the first prediction unit and the second prediction unit by using the first prediction block and the second prediction block.