Intra Prediction Fusion Using Multi-Line References for Low-Complexity Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques face challenges in achieving accurate intra prediction while balancing computational complexity and resource constraints, particularly in resource-constrained devices and applications like streaming.
Innovation Solution
Intra prediction fusion is performed using two or more lines of reference samples to generate a fused predictor, combining reference samples to improve prediction accuracy and reduce computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple lines of reference samples are used for intra prediction fusion, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The current block is divided into multiple prediction blocks, each processed with a specific intra prediction mode. This segmentation allows the system to apply fusion operations selectively to different regions, improving overall prediction accuracy while managing computational complexity through localized processing rather than uniform application across the entire block.
Solution Approach 2:
The patent applies intra prediction fusion selectively based on block characteristics and mode types rather than uniformly to all blocks. By performing fusion operations only where beneficial (partial action), the system improves prediction accuracy for relevant blocks while avoiding unnecessary computational overhead in cases where fusion would provide minimal gain, thus resolving the complexity-accuracy tradeoff.
2Loss of substance
If multiple intra prediction predictors are generated and fused, then residual data is reduced, but coding time increases
Solution Approach 1:
Multiple intra prediction predictors are generated in advance using different reference lines and modes before the actual prediction process. This preliminary generation of multiple predictors allows the system to select and fuse the most appropriate predictors, resulting in better prediction accuracy and smaller residual data. The upfront computation is optimized to reduce overall coding time by avoiding iterative adjustments during the main encoding process.
Solution Approach 2:
The patent generates multiple copies of prediction predictors from different reference lines and modes, then fuses them to create a final prediction. By creating these predictor copies in advance and selecting the best combination, the system reduces residual data more effectively while managing coding time through efficient selection algorithms that avoid exhaustive searching of all possible predictor combinations.
3Reliability
If fusion of predictors from multiple reference lines is performed, then visual quality is improved, but resource requirements increase
Solution Approach 1:
The patent applies different fusion strategies and reference line selections to different regions and blocks based on their local characteristics. By adapting the fusion process to local block properties (such as texture complexity and edge orientations), the system improves visual quality where needed while reducing computational resource requirements in regions where simple prediction suffices, thus resolving the quality-resource tradeoff.
Data Source
AI summary
A video decoder may be configured to determine that a block of video data is encoded using an intra prediction process that utilizes multiple intra prediction predictors; determine a set of reference lines for the intra prediction process; determine a first set of intra prediction predictors based on the set of reference lines; determine a second set of intra prediction predictors based on the set of reference lines; generate a fusion of predictors from the first set of intra prediction predictors and the second set of intra prediction predictors; and decode the block of video data using the fusion of predictors.