Virtual Intra Prediction Mode Selection Using Texture Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining virtual intra prediction modes (VIPMs) in video encoding and decoding are sub-optimal, leading to inefficient transform selection for certain block types, particularly non-angular modes, and do not provide the best performing coding performance.
Innovation Solution
The method involves obtaining virtual intra prediction modes (VIPMs) using prediction samples and template samples from the block's reconstructed neighborhood, employing texture analysis and histogram of gradients (HoG) to select the most accurate VIPM, and refining the selection process through rate-distortion optimization and template-based intra mode derivation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods for determining VIPMs are used, then the encoding and decoding process is simple, but the transform selection accuracy is sub-optimal
Solution Approach 1:
The patent introduces template samples from the reconstructed neighborhood as an intermediary to improve VIPM determination. These template samples serve as a mediator between the current block and the VIPM selection process, enabling more accurate transform selection by comparing prediction samples with template samples from spatially adjacent blocks that have similar texture characteristics.
Solution Approach 2:
The patent replaces simple VIPM determination with a texture analysis-based approach using histogram of gradients (HoG). This substitution transforms the mechanical selection process into a statistical analysis process, where texture patterns are extracted and compared to determine the most suitable VIPM, thereby improving accuracy without requiring complex manual intervention.
2Productivity
If texture analysis and HoG methods are employed to select VIPM, then coding performance is improved, but computational complexity increases
Solution Approach 1:
The patent applies texture analysis and HoG methods selectively rather than universally. By using these computationally intensive techniques only when beneficial for transform selection, the method achieves improved coding performance while avoiding unnecessary computational overhead in cases where simpler VIPM determination would suffice.
Solution Approach 2:
The system performs self-optimization by automatically selecting the appropriate VIPM based on texture characteristics. The texture analysis and HoG computation are performed autonomously to determine the best transform selection, eliminating the need for manual parameter tuning and enabling the system to adapt to different block types and texture patterns independently.
3Measurement precision
If template samples from reconstructed neighborhood are used, then VIPM accuracy is enhanced, but memory requirements increase
Solution Approach 1:
The patent extracts only the necessary template samples from the reconstructed neighborhood that are relevant for VIPM determination. Rather than storing or processing the entire reconstructed neighborhood, the method selectively extracts template samples from spatially adjacent blocks that exhibit similar texture characteristics, thereby reducing memory requirements while maintaining VIPM accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and an apparatus of processing one or more blocks of a picture obtain one or more virtual intra prediction modes, VIPMs, using one or more prediction samples of the block and one or more template samples from a reconstructed neighborhood of the block, or at least a plurality of prediction samples of the block, or one or more prediction samples of the block obtained by an intra prediction mode fusion, IPM-fusion, based prediction.