Chroma Prediction Using Grouped Reference Values and Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as H.265/HEVC, face challenges in accurately predicting chroma components, leading to low prediction accuracy and inefficient coding rates, especially for high-definition and ultra-high-definition video applications like VR videos.
Innovation Solution
The proposed method improves chroma component prediction by grouping neighboring reference values and establishing multiple computing models based on thresholds, allowing for more accurate prediction of chroma values through Cross-Component Linear Model Prediction (CCLM) in the next-generation video coding standard H.266, which includes Cross-Component Prediction (CCP) and Cross-Component Linear Model (CCLM).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only neighboring reference luma values and neighboring reference chroma values are considered for computing linear model parameters, then the computation is simple, but the prediction accuracy of chroma component is low
Solution Approach 1:
The patent divides the prediction process into multiple segments: first computing a preliminary linear model using neighboring reference values, then computing a correction model using reconstructed luma values, and finally combining both models to obtain the final chroma prediction. This segmentation allows each model to focus on specific aspects of prediction, improving overall accuracy while managing complexity.
Solution Approach 2:
The patent introduces reconstructed luma values as an intermediary element that bridges the simple neighboring reference value approach and the final accurate prediction. The correction model uses these reconstructed values to adjust the preliminary prediction, acting as a mediator that enhances accuracy without requiring complete redesign of the prediction system.
2Measurement precision
If a simple linear model is used for chroma prediction, then the coding complexity is low, but the prediction accuracy is insufficient for high-definition and ultra-high-definition videos
Solution Approach 1:
The patent implements a dynamic two-stage prediction system where the first stage uses a simple linear model for quick computation, and the second stage dynamically applies correction based on reconstructed luma values when prediction accuracy is insufficient. This dynamic approach adapts the complexity of computation to the actual prediction needs, balancing accuracy and efficiency.
Solution Approach 2:
The patent changes the parameters used in prediction from only neighboring reference values to include both neighboring reference values and reconstructed luma values. By modifying the input parameters of the prediction model, the system achieves higher prediction accuracy for HD and UHD videos while maintaining reasonable coding complexity through efficient parameter utilization.
3Measurement precision
If multiple computing models are established based on grouping neighboring reference values, then the prediction accuracy is improved, but the computation complexity increases
Solution Approach 1:
The patent segments the computation into two distinct models: a preliminary linear prediction model and a correction model. Each model processes different aspects of the prediction task, allowing the system to achieve high accuracy through specialized sub-models rather than one complex monolithic model, thereby managing computational complexity more effectively.
Solution Approach 2:
The patent applies different prediction strategies to different parts of the prediction process: the preliminary model handles the base prediction using neighboring reference values, while the correction model specifically addresses prediction errors using reconstructed luma values. This local specialization of prediction quality in different computational stages improves overall accuracy while keeping each stage's complexity manageable.
Data Source
Figure 1A~1C
Figure 2A~2B
Figure 3A~3C
AI summary
A video image component prediction method and device, and a computer storage medium. The method comprises: acquiring at least one threshold value according to first image component reconstruction values of all sampling points of a current coding block (S601); carrying out grouping according to a comparison result of a first image component neighboring reference value of the current coding block and the at least one threshold value, so as to obtain at least two groups of first image component neighboring reference values and second image component neighboring reference values (S602); establishing at least two computing models according to the at least two groups of first image component neighboring reference values and second image component neighboring reference values (S603); selecting, from the at least two computing models, a computing model corresponding to each sampling point in the current coding block according to a comparison result of the first image component reconstruction values of all the sampling points of the current coding block and the at least one threshold value (S604); and according to the computing model corresponding to each sampling point in the current coding block, acquiring a second image component prediction value of each sampling point in the current coding block (S605).