Kernel Density Estimation for Robust Gray Patch White Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automatic white balancing techniques, such as the Gray World method, often fail to accurately normalize images due to over-biasing from dominant colors or large surfaces, leading to inaccurate hue representation and distortion, especially in scenes with human skin or blue skies, and can result in unrealistic color corrections.
Innovation Solution
A method for robust detection of gray patches in images, using kernel density estimation to identify candidate gray points, refine estimates by compensating for outliers, and dynamically adjust thresholds for color preference and tuning, ultimately converting the refined gray point into white balance gains for improved color temperature control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the Gray World method averages all pixel colors to estimate illuminant color, then the method is simple to implement and computationally efficient, but it produces poor white balance performance in scenes with dominant colors or large colored surfaces
Solution Approach 1:
The patent applies local quality by assigning different weights to different regions of the image based on their likelihood of containing gray surfaces. Instead of treating all pixels equally, the algorithm identifies and weights specific local regions (e.g., shadow regions, non-saturated areas) more heavily in the illuminant estimation process, thereby improving accuracy in scenes with dominant colors while maintaining computational efficiency.
Solution Approach 2:
The patent changes the parameter of pixel weighting from uniform (Gray World) to adaptive based on local image characteristics. By introducing weight factors that vary across different image regions based on saturation, luminance, and other local properties, the method transforms the simple average into a weighted average that adapts to local conditions, resolving the contradiction between simplicity and accuracy.
2Ease of operation
If the Gray World method uses the average color of the entire image, then the implementation is straightforward, but it causes gray objects to be rendered inaccurately and produces distorted coloring in certain scenes
Solution Approach 1:
The patent implements local quality by differentiating between different image regions and applying appropriate weighting. It identifies regions likely to contain gray surfaces (such as shadow regions and non-saturated areas) and gives them higher weights, while reducing the influence of colored regions. This localized approach maintains implementation simplicity while significantly improving color accuracy and reliability.
Solution Approach 2:
The patent introduces intermediary weight factors that mediate between the simple average calculation and the final illuminant estimate. These weight factors act as intermediaries that filter out problematic regions (saturated colors, highlights, shadows) while preserving the computational efficiency of the average-based approach, thereby maintaining ease of operation while improving reliability.
3Measurement precision
If conventional white balancing methods apply strong bias correction to counteract dominant colors, then they can restore some color balance, but they cause over-graying of skin tones and unintended hue shifts in other objects
Solution Approach 1:
The patent applies local quality by treating different image regions differently based on their characteristics. Instead of applying a uniform bias correction, it identifies skin-tone regions and other sensitive areas and applies reduced or modified correction in those regions, while applying stronger correction in regions with dominant colors. This prevents over-graying of skin tones and unintended hue shifts while still achieving color normalization in problematic areas.
Solution Approach 2:
The patent applies preliminary anti-action by preemptively identifying regions that would be adversely affected by strong bias correction (such as skin tones and naturally colored objects) and adjusting the correction strategy before applying it. By pre-weighting or pre-identifying sensitive regions, the method prevents harmful over-correction while maintaining effective normalization in regions where it is needed.
Data Source
AI summary
Embodiments of the present invention are directed to methods and systems for robust weighting of gray patches in automatic white balancing in an image-capture device by utilizing kernel density estimation techniques with dynamically variable bandwidth to determine the probability density of samples to create an initial estimate, then verifying the initial gray point estimate to account for outliers. In one embodiment, given a set of image data, an initial gray point estimate in a color space is determined for the set of image data. The initial estimate is then refined by weighting the sub-population with the greatest probability of being gray. A final evaluation that includes a further comparison to pre-programmed constraints determines a final estimate, which can still be further tuned according to user preferences by adjusting color biases. The resulting final gray point estimate provides greater stability, and greatly improved accuracy over traditional techniques and solutions.