Video Intra-Prediction Using Reduced Template Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs like HEVC and VVC face challenges in maintaining coding efficiency while reducing complexity when using Decoder-side Intra Mode Derivation (DIMD) for luma and chroma samples, particularly due to increased computational resources.
Innovation Solution
A method for deriving intra-prediction modes by analyzing gradients of samples in a template area, calculating a histogram of gradients, and selecting at most two angular intra-prediction modes to determine the intra-prediction mode, with filtering windows centered at reduced middle line sample positions to reduce resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Decoder-side Intra Mode Derivation (DIMD) is used for luma and chroma samples, then coding efficiency is improved, but computational complexity increases
Solution Approach 1:
The patent segments the template area into multiple regions and processes only selected regions (e.g., every other line or specific zones) rather than all samples. This segmentation approach reduces the number of gradient calculations while preserving essential directional information, thereby lowering computational complexity by up to 20% at the encoder and 5% at the decoder while maintaining coding efficiency.
Solution Approach 2:
The patent applies partial action by calculating histograms of gradients for only a subset of template area samples rather than all samples. By selectively processing middle line sample positions with filtering windows centered at reduced positions, the method achieves sufficient prediction accuracy with reduced computational effort, resolving the contradiction between coding efficiency and computational complexity.
2Measurement precision
If filtering windows are centered at all middle line sample positions, then prediction accuracy is improved, but resource requirements increase
Solution Approach 1:
The patent extracts and processes only the most relevant samples from the template area by centering filtering windows at selected middle line sample positions rather than all positions. This extraction approach identifies and utilizes the critical gradient information needed for accurate prediction while discarding redundant calculations, thereby reducing resource requirements while maintaining prediction accuracy.
Solution Approach 2:
The patent applies local quality by differentiating the processing intensity across different regions of the template area. Instead of uniform processing of all samples, the method applies filtering windows selectively at specific middle line positions where gradient information is most valuable, optimizing the balance between prediction accuracy and computational resources.
Data Source
AI summary
The present disclosure relates to encoding/decoding a block of samples of a video picture in which a DIMD mode is derived for luma and chroma samples of a sample block to be predicted by filtering samples of at least one template area, said filtering using filtering windows centered at middle line sample positions of the at least one template area filtering samples in at least one shaped template area. An integer number of middle line sample positions of the at least one template area at which the filtering windows are centered, is lower than a total integer number of middle line sample positions of the at least one template area.