Cross-component prediction model for video chroma blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cross-component prediction models in video compression are inefficient due to cross-component content inconsistencies, particularly when local texture or lighting changes occur, leading to sub-optimal predictions and coding inefficiencies.
Innovation Solution
A method that involves analyzing an analysis template for cross-component prediction, detecting variations indicating inconsistencies, splitting reference samples into subsets, and applying weighted values based on content consistency to derive improved cross-component prediction models for chroma blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If cross-component prediction model uses L-shaped template around current block, then prediction can be performed with simple structure, but prediction accuracy deteriorates due to cross-component content inconsistencies in the template
Solution Approach 1:
The patent segments the L-shaped template into multiple sub-templates based on detected variations. Instead of using the entire template uniformly, the method divides it into regions with consistent content characteristics, allowing accurate prediction only from reliable sub-templates while excluding inconsistent regions that would degrade prediction quality.
Solution Approach 2:
The patent applies local quality by assigning different weights to different parts of the template based on their content consistency. Regions with high consistency receive higher weights for prediction, while regions with low consistency (containing object borders or texture variations) receive lower weights or are excluded, thereby improving overall prediction accuracy without requiring complete template rejection.
2Productivity
If cross-component prediction is applied to all reference samples in the template, then coding efficiency is maintained with uniform processing, but prediction performance deteriorates when content inconsistencies are present
Solution Approach 1:
The patent introduces dynamic adaptation by detecting variations in the template and adjusting the prediction strategy accordingly. The method dynamically identifies consistent regions and adapts the prediction process to use only those regions, transforming a static uniform processing approach into a dynamic content-aware approach that maintains reliability while preserving coding efficiency.
Solution Approach 2:
The patent changes the parameter of template usage from uniform inclusion to selective weighting based on content consistency. By introducing consistency metrics and adjusting the weight parameters of different template regions, the method optimizes prediction performance without fundamentally changing the cross-component prediction framework, thereby maintaining coding efficiency while improving reliability.
3Ease of operation
If the prediction model relies on pre-defined neighborhood template, then implementation is simple and fast, but it introduces blindness to content inconsistencies causing sub-optimal prediction
Solution Approach 1:
The patent applies preliminary action by performing variation detection on the template before executing the prediction process. This preliminary analysis identifies consistent and inconsistent regions in advance, allowing the prediction to be performed only on reliable data. This approach maintains implementation simplicity by adding a pre-processing step rather than fundamentally changing the prediction mechanism.
Solution Approach 2:
The patent introduces feedback by using the detected content consistency information to guide the prediction process. The variation detection results provide feedback on which template regions are reliable, and this feedback is used to adjust the prediction weighting and selection, creating a closed-loop system that improves prediction optimality while maintaining ease of implementation through systematic feedback utilization.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The present disclosure relates to the cross-component prediction method of a chroma block of a video picture. Cross-component prediction models are derived based on weighted reference samples in a model template defined around a co-located luma block of the chroma block to be predicted. An analysis template analysis template comprises reconstructed luma samples of a co-located luma block of the chroma block and the reference samples of a model template. The reference samples of the model template are split into at least two sub-sets of reference samples of the analysis template based on at least one detected variation in the reference samples of the analysis template. The weighted reference samples in the model template are obtained by weighing the reference samples of sub-sets of the analysis template that belong to the model template by weighting values, each weighting value being determined based on the content consistency between the reference samples of a sub-set of reference samples of the analysis template belonging to the model template and the reconstructed samples of the co-located luma block of the analysis template.