Multi-Reference Intra Prediction for Uneven Texture Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-reference line intra prediction methods assume uniform texture distribution on the upper and left sides of a coding unit, leading to inaccurate prediction values when texture distributions are uneven, resulting in lower prediction accuracy.
Innovation Solution
A multi-reference line intra prediction method that allows for different distances between the target reference row and column and the current coding unit, adapting to non-uniform texture distributions by determining a target intra prediction mode, reference row, and reference column based on intra prediction mode information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If multi-reference line intra prediction methods assume uniform texture distribution on upper and left sides of coding unit, then device complexity is reduced, but prediction accuracy deteriorates when texture distributions are uneven
Solution Approach 1:
The patent applies local quality by allowing different reference rows and columns to be selected for different regions of the coding unit based on local texture characteristics. The decoding device determines target reference rows and columns that can differ from each other, enabling each region to use reference lines most suitable for its local texture distribution pattern, thereby improving prediction accuracy without requiring complex uniform assumptions across the entire block.
Solution Approach 2:
The patent implements dynamics by making the reference line selection adaptive rather than fixed. The intra prediction mode information dynamically indicates which target reference row and target reference column to use based on the actual texture distribution observed in the video content. This dynamic selection allows the prediction method to adapt to varying texture patterns in different coding units and different regions within coding units.
2Measurement precision
If different reference rows and columns at different distances are used, then prediction accuracy is improved for non-uniform textures, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-defining a set of candidate reference rows and columns at different distances from the coding unit. These candidate reference lines are prepared in advance according to the angular prediction mode, and the decoding device only needs to select from this pre-organized set based on the intra prediction mode information. This preliminary organization reduces the complexity of real-time selection while enabling access to multiple reference lines at different distances.
Solution Approach 2:
The patent uses parameter changes by varying the distance parameters of reference rows and columns based on the intra prediction mode. The target reference row and target reference column are selected from multiple candidates at different distances, allowing the system to change the reference line distance parameter to match the texture distribution characteristics. This parameter variation enables accurate prediction for non-uniform textures without requiring complex processing, as it simply involves selecting from pre-defined distance options.
Data Source
AI summary
This application discloses a multi-reference line intra prediction method and apparatus, an electronic device, and a readable storage medium. The multi-reference line intra prediction method of embodiments of this application includes: obtaining, by a decoding end, intra prediction mode information of a current coding unit from a bitstream, where the intra prediction mode information is used to indicate a target intra prediction mode, a target reference row, and a target reference column; determining, by the decoding end based on the intra prediction mode information, a target intra prediction mode, a target reference row, and a target reference column that are used by the current coding unit; and determining, by the decoding end, a prediction value of the current coding unit based on the target intra prediction mode, the target reference row, and the target reference column.