Intra Prediction Direction Derivation for Accurate Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for constructing intra prediction directions in video coding, such as incrementing or decrementing the index number of a known candidate prediction direction by 1, do not provide a sufficient accurate prediction effect during intra prediction.
Innovation Solution
Perform vector operations on multiple first prediction directions to obtain second prediction directions, combining them with the first prediction modes to form a second prediction mode set for intra prediction on the current processing block.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vector operations are performed on multiple first prediction directions to obtain second prediction directions, then the accuracy of intra prediction is improved, but the device complexity increases
Solution Approach 1:
The patent transforms prediction directions from simple index number increments to vector operations on directional vectors. Each prediction direction is represented as a vector with magnitude and direction, allowing for more accurate derivation of candidate prediction directions through vector addition and scaling operations on neighboring block prediction vectors.
Solution Approach 2:
The patent introduces dynamic adaptation by deriving prediction directions based on the actual directional characteristics of neighboring blocks rather than using fixed index-based increments. The vector operations allow the prediction system to adapt to the local directional patterns in the video content, making the prediction more accurate for different block types and orientations.
2Productivity
If multiple candidate prediction modes are selected for intra prediction, then the compression ratio is improved, but the number of bits required for representing the prediction mode increases
Solution Approach 1:
The patent performs preliminary derivation of candidate prediction directions from neighboring blocks before the actual prediction process. By pre-computing the most likely prediction directions based on neighboring block characteristics, the system narrows down the candidate set to only the most relevant directions, reducing the number of bits needed to signal the final prediction mode while maintaining high compression efficiency.
Data Source
AI summary
A method for intra prediction includes: obtaining multiple previously reconstructed neighbouring blocks corresponding to a current processing block; determining prediction modes, that are signalled in a bitstream, corresponding to neighbouring blocks of the multiple previously reconstructed neighbouring blocks, to obtain multiple first prediction modes; if the multiple first prediction modes comprise at least two directional modes, taking directional modes comprised in the multiple first prediction modes as first prediction directions; performing, according to a preset operation rule, operation on multiple first prediction directions of the first prediction directions to obtain second prediction directions; obtaining a prediction mode set according to the second prediction directions and the multiple first prediction modes; and performing intra prediction on the current processing block based on the prediction mode set.