Chroma Intra Prediction Using Cross-Component Linear Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as ITU-T H.265 and JVET-Q2001, face challenges in efficiently performing intra prediction, particularly for chroma components, especially when handling cases like unavailable neighboring samples and 4:2:2 chroma formats, which can lead to suboptimal prediction results.
Innovation Solution
The implementation of a cross-component linear model (CCLM) for chroma intra-prediction, which determines the number of available neighboring chroma samples and uses a specific prediction process to generate prediction samples, including down-sampling and filtering techniques, to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional intra prediction methods are used for chroma components, then the coding process is simple, but prediction accuracy deteriorates when neighboring samples are unavailable or in 4:2:2 chroma formats
Solution Approach 1:
The patent introduces luma samples as an intermediary to predict chroma samples when direct chroma neighboring samples are unavailable. The cross-component linear model uses luma samples (which are always available) as a mediator to derive chroma prediction values through linear transformation, resolving the contradiction between maintaining simplicity and improving accuracy in cases with limited chroma reference samples.
Solution Approach 2:
The patent changes the prediction parameter from direct chroma sample copying to a linear transformation model that incorporates luma sample values. By transforming the prediction approach from spatial copying to cross-component linear modeling, the system achieves better prediction accuracy while managing complexity through standardized transformation processes.
2Measurement precision
If cross-component linear model with down-sampling is used, then prediction accuracy for chroma improves, but computational complexity increases
Solution Approach 1:
The patent applies partial down-sampling of luma samples rather than processing all luma samples at full resolution. By selectively using down-sampled luma samples for the linear transformation, the method achieves sufficient prediction accuracy while reducing the computational power required compared to using full-resolution luma samples for all predictions.
3Measurement precision
If more neighboring samples are used for prediction, then prediction quality improves, but the complexity of handling unavailable samples increases
Solution Approach 1:
The patent creates a universal prediction framework where luma samples serve multiple functions: they act as direct predictors when chroma samples are unavailable, and they serve as the basis for cross-component linear transformation. This multi-functional use of luma samples simplifies the handling of various sample availability scenarios compared to maintaining separate prediction paths for different cases.
Data Source
AI summary
A method of generating predicted samples for a current transform block in video coding is disclosed. The method comprising: setting a neighboring left luma sample value defined with x equal to −1, −2 or −3, corresponding to the current transform block, being equal to a reconstructed luma sample; setting a neighboring top-left luma sample value defined with x equal to −2 and y equal to −1, corresponding to the current transform block, being equal to the neighboring left luma sample value, in a case that the neighboring top-left luma sample value is unavailable; deriving a down-sampled neighboring left luma sample value for the current transform block by using the neighboring left luma sample value and on a chroma format; and generating the prediction samples by using the down-sampled neighboring luma sample values.