Combined Video Prediction With Regression-Based Mode Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like VVC face challenges in optimizing the prediction accuracy and efficiency of intra and inter prediction modes, particularly in handling complex video content and maintaining decoder complexity within manageable limits.
Innovation Solution
A method and apparatus for video coding that utilizes a regression-based technique to derive a combined predictor by weighting multiple source terms, including inter and intra prediction modes, using templates to improve prediction accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple prediction modes are combined to improve prediction accuracy, then coding performance is improved, but decoder complexity increases
Solution Approach 1:
The patent changes the parameter of weighting method from simple fixed weighting to regression-based adaptive weighting. By using regression techniques to dynamically determine optimal weights for different prediction modes based on local video content characteristics, the system achieves higher prediction accuracy without proportionally increasing decoder complexity, as the regression model can be efficiently implemented.
Solution Approach 2:
The patent introduces dynamic weight adjustment through regression-based techniques. Instead of using static predetermined weights, the system dynamically adapts the weighting of different prediction modes (intra, inter, palette) based on local content characteristics analyzed through regression modeling. This dynamic approach improves prediction accuracy for diverse video content while maintaining manageable complexity through efficient regression computation.
2Productivity
If regression-based weighting is applied to combine prediction modes, then coding efficiency is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial regression-based weighting rather than to all prediction scenarios. The system selectively uses regression techniques for specific prediction units or regions where they provide the most benefit, rather than universally applying complex regression modeling throughout the entire video stream. This partial application maintains coding efficiency improvements while limiting the increase in computational complexity to only where necessary.
3Device complexity
If traditional fixed weighting is used for combining prediction modes, then decoder complexity is kept low, but prediction accuracy is insufficient
Solution Approach 1:
The patent transforms the weighting parameter from fixed predetermined values to dynamically determined regression-based weights. This parameter change enables the system to adaptively optimize the combination of intra, inter, and palette prediction modes according to local content characteristics, significantly improving prediction accuracy while the regression implementation maintains acceptable complexity levels.
Data Source
AI summary
A method and apparatus for video coding using combined prediction. According to this method, a combined predictor is derived, wherein the combined predictor comprises a weighted combination of one or more source terms. One or more weightings for the weighted combination of one or more source terms are derived using a regression technique based on relationship estimation based on one or more templates. The one or more source terms correspond to different prediction modes for the current block and the different prediction modes comprise at least one non-intra mode. Or, the one or more source terms correspond to member samples in a pattern determined associated with a to-be-predicted sample for the current block and each member sample corresponds to an initial predicted sample of combining prediction mode. The current block is encoded or decoded using information comprising the combined predictor.