Chroma Prediction Weighting for Cross-Component Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video coding standards like H.266/VVC, the prediction accuracy of chroma values is reduced due to deviations in neighboring reference luma and chroma values, leading to inaccuracies in the linear model used for cross-component prediction.
Innovation Solution
A method for predicting video color components that considers correlations and similarities between neighboring reference and reconstructed values to determine weight coefficients, adjusting the linear model to improve accuracy by allocating different weights to neighboring samples based on their correlation with the current coding block.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a linear model is constructed using neighboring reference luma and chroma values, then chroma prediction can be performed, but prediction accuracy is reduced when reference values deviate greatly from current coding block parameters
Solution Approach 1:
The patent applies local quality by differentiating between different types of neighboring reference samples based on their correlation with the current coding block. It identifies and weights samples locally according to their reliability, giving higher weights to samples with similar characteristics (luma or chroma) and lower weights to deviated samples, thereby improving prediction accuracy adaptively.
Solution Approach 2:
The patent changes the parameter of weight coefficients dynamically based on the correlation between reference samples and current coding block. By adjusting weight coefficients according to sample correlation (using metrics like sum of absolute differences), the linear model adapts to local characteristics, resolving the contradiction between using reference samples and maintaining accuracy.
2Ease of manufacture
If all neighboring reference samples are used equally in linear model construction, then calculation is simplified, but prediction accuracy deteriorates when some samples are greatly deviated
Solution Approach 1:
Instead of treating all reference samples uniformly, the patent applies local quality by assigning different weight coefficients to different samples based on their local correlation with the current coding block. This differentiated approach maintains calculation feasibility while significantly improving prediction accuracy by excluding or down-weighting deviated samples.
Solution Approach 2:
The patent implements partial action by selectively using only the most correlated reference samples (or giving them higher weights) rather than all samples equally. This partial selection of high-quality reference samples improves accuracy without requiring complex processing of all samples, balancing simplicity and precision.
Data Source
AI summary
A method for reading a bitstream, a method for storing a bitstream and a method for transmitting a bitstream are provided. The method includes: a bitstream is read, and the following operations are executed to decode the bitstream: obtaining a luma component neighboring reference value and a luma component reconstructed value corresponding to a current block, wherein the luma component neighboring reference value represents a luma component parameter corresponding to a neighboring reference sample of the current block; determining a weight coefficient corresponding to the neighboring reference sample; determining a scaling factor according to the weight coefficient and the luma component neighboring reference value; and obtaining a chroma component predicted value corresponding to the current block based on a linear model corresponding to the scaling factor.