White Point Detection via Confidence Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional white balance methods, such as the Retinex method, are inadequate in resisting noise and handling over-exposed pixels in images, leading to suboptimal image processing.
Innovation Solution
A method and system for white point detection that determines pixel candidates based on location and brightness, estimating confidences to identify reliable white points by filtering out noise and over-exposed pixels through a confidence conversion process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the Retinex method is used to detect white points, then the detection process is simple and fast, but the method is too simple to resist against noises and suffers from over-exposed pixels
Solution Approach 1:
The patent segments the image into multiple local regions and performs white point detection independently in each region. This segmentation allows the method to handle local variations and noise more effectively while maintaining computational efficiency. Each region's white point candidates are evaluated separately, reducing the impact of noise from other areas.
Solution Approach 2:
The patent introduces confidence values as an intermediary metric to evaluate white point candidates. Instead of directly selecting pixels with maximum channel values, the method computes confidence values based on local color distribution and uses these as a mediator to filter out unreliable candidates caused by noise or over-exposure before final selection.
2Device complexity
If the Retinex method is used to detect white points, then the detection process is simple and fast, but the method suffers from over-exposed pixels in the image
Solution Approach 1:
The patent converts the harmful effect of over-exposed pixels into a beneficial filtering process. By computing confidence values that reflect local color distribution characteristics, over-exposed pixels naturally receive low confidence scores and are automatically filtered out. This transforms what was previously a harmful artifact into a useful indicator for eliminating false candidates.
Solution Approach 2:
The patent changes the selection parameter from simple channel value comparison to a composite confidence metric that incorporates local color distribution information. This parameter transformation enables the method to distinguish between true white points and over-exposed pixels by evaluating multiple aspects of local color characteristics rather than relying on a single intensity value.
3Ease of operation
If conventional white balance methods are used, then the adjustment process is straightforward, but the methods are inadequate in resisting noise
Solution Approach 1:
The patent divides the image into multiple local regions and performs confidence-based filtering in each region independently. This segmentation approach maintains operational simplicity while improving noise resistance, as each region can be processed with the same straightforward confidence computation and filtering logic.
Solution Approach 2:
The patent introduces a feedback mechanism where confidence values are computed based on local color distribution and used to iteratively refine the selection of white point candidates. This feedback loop allows the method to automatically adjust and reject unreliable candidates without requiring complex manual intervention, maintaining ease of operation while significantly improving noise resistance.
Data Source
AI summary
A method of white point detection for an image is disclosed. The method includes determining a plurality of pixels of the image as a plurality of white point candidates; estimating a plurality of candidate confidences according to a plurality of brightness values corresponding to the plurality of white point candidates; and determining a plurality of white points of the image according to the plurality of candidate confidences.