Cross-Component Intra Prediction With Adaptive Reference Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression techniques, such as VVC, rely on preset reference samples for cross-component prediction, which is inefficient and does not fully leverage spatial and statistical information of luma and chroma components, leading to suboptimal video encoding efficiency and quality.
Innovation Solution
A method and apparatus that derive a cross-component relation model by selecting optimal reference samples based on spatial and statistical information of luma components and neighboring samples for chroma block prediction, adjusting temporary sampling positions to enhance video coding efficiency and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If preset reference samples are used for cross-component prediction, then the encoding process is simple, but the video encoding efficiency and quality are suboptimal
Solution Approach 1:
The patent performs preliminary actions by deriving temporary sampling positions from neighboring samples before final position selection. This preliminary derivation uses characteristics of reconstructed luma component or neighboring samples to establish initial candidate positions, which are then refined through evaluation and adjustment to achieve optimal reference sample selection for cross-component prediction
Solution Approach 2:
The patent implements dynamic reference sample selection by evaluating and adjusting temporary sampling positions based on actual luma and chroma characteristics. Instead of using fixed preset positions, the method dynamically determines final sampling positions by assessing the suitability of temporary positions and adjusting them to maximize prediction accuracy for the current block
2Measurement precision
If preset reference samples are used for cross-component prediction, then the processing is straightforward, but spatial and statistical information of luma and chroma components is not fully leveraged
Solution Approach 1:
The patent applies local quality by deriving sampling positions specific to each current block based on its local characteristics. The temporary sampling positions are derived from neighboring samples of the current block, and final positions are selected by evaluating local luma and chroma characteristics, ensuring that each block uses reference samples optimized for its specific spatial and statistical properties
Solution Approach 2:
The patent implements feedback mechanisms by evaluating temporary sampling positions based on the actual characteristics of luma and chroma components. The evaluation process provides feedback on the suitability of temporary positions, which is then used to adjust and select final sampling positions, creating a closed-loop system that continuously optimizes reference sample selection based on observed data
Data Source
AI summary
A method is disclosed for selecting a reference sample for deriving a cross-component relation model in intra prediction. In the disclosed embodiments, a video decoding device derives temporary sampling positions from neighboring samples of the current block by using characteristics of a reconstructed luma component or neighboring samples of the current block. The video decoding device evaluates and adjusts the temporary sampling positions to select final positions and derives the cross-component relation model by using samples of the final positions. The video decoding device generates chroma prediction values of the chroma component by applying the cross-component relation model to the luma component.