Template-Based Intra Prediction Split for Varying Block Directionality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing backward intra prediction modes lack a solution for determining the best mode split of a block with varying directionality, leading to inefficiencies in video coding.
Innovation Solution
A method for determining a final seed intra prediction mode and mode split by minimizing distortion metrics using neighbor templates, allowing for multiple predictions based on reconstructed neighborhoods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If backward estimation methods are used to determine multiple prediction modes, then prediction accuracy is improved, but the ability to determine optimal mode split for blocks with varying directionality deteriorates
Solution Approach 1:
The current block is divided into multiple partitions based on the determined mode split, allowing different prediction modes to be applied to different partitions. This segmentation enables the system to handle blocks with varying directionality by treating each partition independently with its own optimal prediction mode.
Solution Approach 2:
The method performs preliminary determination of the mode split and seed intra prediction mode before generating the final prediction. By using neighbor templates and distortion metrics in advance, the system pre-determines the optimal partitioning strategy, which then guides the application of multiple prediction modes to different partitions.
2Measurement precision
If multiple predictions are applied to the same block, then prediction quality is improved, but signaling overhead increases
Solution Approach 1:
Different prediction modes are applied to different partitions of the block based on local characteristics. The mode split and seed intra prediction mode are determined to match local directionalities within the block, allowing each partition to use the most appropriate prediction mode for its specific region while avoiding unnecessary signaling for the entire block.
Solution Approach 2:
The method applies multiple predictions only where necessary - specifically to partitions that benefit from different directionalities. By determining the mode split based on neighbor templates and distortion metrics, the system applies partial action (multiple predictions) only to the extent needed to capture local variations, rather than applying multiple predictions uniformly across the entire block.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for comprising: obtaining (1001) a current block; obtaining (1002) a neighbor template of samples in a reconstructed neighborhood of the current block; determining (1004) a seed intra prediction mode using a minimization of a distortion metric based on predictions of the neighbor template from reference samples in a reconstructed neighborhood of the neighbor template; determining (1006) a mode split for the current block based on the seed intra prediction mode; and, generating a predictor of the current block by applying (1012) an intra prediction to each partition of the current block specified by the mode split, the intra prediction being based on the seed intra prediction mode.