Color Component Prediction Using Filtered Reference Sample Subsets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing color prediction models in video coding and decoding require a large number of samples for model construction, leading to high calculation complexity and memory bandwidth, and are prone to inaccuracies due to exceptional samples.
Innovation Solution
Reduce the number of pixels in the reference pixel set by screening out unimportant or exceptional reference pixels, using a selected subset to calculate the model parameter for cross-component prediction processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of samples are used to construct the prediction model, then the model parameter accuracy may be improved, but the calculation complexity and memory bandwidth increase significantly
Solution Approach 1:
The patent extracts only the necessary and effective reference pixels from the complete reference pixel set to form a subset. This is achieved by identifying and removing unimportant or exceptional reference pixels that do not contribute meaningfully to the prediction model, thereby reducing the number of samples needed while maintaining or improving model accuracy and reducing calculation complexity.
Solution Approach 2:
The patent applies local quality by treating different reference pixels differently based on their importance. Instead of uniformly using all reference pixels, the method identifies specific regions or individual pixels within the reference pixel set that have higher quality or importance, and selectively uses those for model construction, thereby improving efficiency without sacrificing accuracy.
2Reliability
If a large number of samples are used to construct the prediction model, then more comprehensive data coverage is achieved, but the memory bandwidth requirement increases
Solution Approach 1:
The patent extracts a representative subset of reference pixels that maintains data coverage reliability while reducing the total quantity of data that needs to be processed. By carefully selecting which pixels to include in the subset, the method ensures that the essential characteristics and variations in the reference data are preserved, achieving reliable prediction without requiring high memory bandwidth for processing all original samples.
3Productivity
If all reference pixels are used in the prediction model, then complete data utilization is achieved, but exceptional samples reduce the model accuracy
Solution Approach 1:
The patent removes exceptional or unimportant reference pixels from the complete reference pixel set to form a cleaned subset. This extraction process identifies and eliminates outliers or pixels with abnormal characteristics that would negatively impact model accuracy, while retaining the majority of useful reference data, thereby achieving both good data utilization and high prediction accuracy.
Solution Approach 2:
The patent applies local quality by evaluating each reference pixel's contribution to the prediction model and selectively including only those with high quality. This approach ensures that exceptional samples are excluded while maintaining comprehensive utilization of the remaining valid reference pixels, optimizing both data utilization efficiency and model accuracy.
Data Source
AI summary
Provided are a method for predicting a colour component, an encoder and a decoder. The method includes: a first reference sample set of a colour component to be predicted of a current block is determined; a reference sample subset is determined from the first reference sample set, where the reference sample subset includes one or more candidate samples selected from the first reference sample set; and a model parameter of a prediction model is calculated according to positions of reference samples in the reference sample subset, where the prediction model is configured to perform prediction processing on the colour component to be predicted of the current block.