Automatic White Balancing Using Illuminant Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automatic white balancing methods, such as the Gray World approach, often result in sub-optimal illuminant color estimation, leading to distorted images due to biased estimates from large colored surfaces and random sensor noise, which can incorrectly neutralize colors in scenes.
Innovation Solution
The method projects the initial illuminant color estimate onto a plot of common illuminants to find a more optimal color, adjusting the image data by the reciprocal of the closest point on this plot, thereby improving color normalization and reducing user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If the Gray World method is used to estimate illuminant color by averaging pixel values, then the process is simple and automated, but the estimated illuminant color becomes biased by large colored surfaces and sensor noise, resulting in distorted images
Solution Approach 1:
The patent introduces an intermediary constraint based on physical illuminant characteristics (Planckian locus for thermal radiators and spectral lines for gas discharge lamps) that mediates between the raw pixel average and the final illuminant estimate. This intermediary prevents direct use of biased averages while maintaining automated operation, resolving the contradiction between automation and accuracy.
Solution Approach 2:
The patent changes the parameter space by constraining the illuminant estimate to lie on physically valid illuminant curves in color space rather than allowing free variation. This parameter transformation ensures that automated estimation produces only physically plausible results, improving accuracy without sacrificing automation.
2Quantity of substance
If pixels from large colored surfaces are included in the average calculation, then more image data is utilized, but the illuminant estimate becomes overly biased toward the dominant color, causing incorrect neutralization
Solution Approach 1:
The physical illuminant constraint acts as an intermediary that allows utilization of all pixel data while preventing domination by large colored surfaces. The constraint ensures that even if the raw average is heavily biased, the final estimate remains on the valid illuminant curve, resolving the contradiction between using more data and maintaining accuracy.
3Productivity
If random sensor noise pixels are given weight during illuminant determination, then all available data is used, but the estimate becomes biased by these random colored pixels
Solution Approach 1:
The physical illuminant constraint serves as a mediator that allows efficient processing of all pixels while filtering out the influence of random noise. By projecting the raw average onto the illuminant curve, the system efficiently uses all data without being unduly influenced by noise pixels, resolving the contradiction between productivity and precision.
4Device complexity
If the average color of the image is used as the estimated illuminant color, then the calculation is straightforward, but the result is sub-optimal for normalization and can produce distorted images
Solution Approach 1:
The patent introduces a simple intermediary step of projecting the raw average onto the physical illuminant curve. This adds minimal complexity while dramatically improving color normalization accuracy by ensuring the estimate conforms to physical reality, resolving the contradiction between simplicity and precision.
Solution Approach 2:
The patent transforms the parameter representation by constraining the illuminant to lie on physically valid curves in color space. This parameter change maintains computational simplicity while improving normalization accuracy by eliminating unphysical estimates.
Data Source
AI summary
Embodiments of the claimed subject matter are directed to methods for automatic white balancing in an image-capture device. In one embodiment, given an estimated illuminant color (e.g., derived from the Gray World method), a more optimal illuminant color can be found by projecting this point to a plot of common illuminants to determine the closest point on the plot of common illuminants. Once the closest point of the plot of common illuminants is derived, the actual image (e.g., pixel) data of the scene is adjusted by the value of the closest point on the plot of common illuminants so that the light is normalized for the scene.