Auto White Balance Correction Using Face Area Feature Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing white balance adjustment methods using face detection fail to accurately identify the light source when detecting non-human faces, such as sculptures, or painted faces without skin color, leading to erroneous white balance corrections.
Innovation Solution
An auto white balance correction method that calculates a total correction value by comparing and combining normal and face-specific correction values based on feature data from the entire image and the face area, using weighted averages and differences in light source colors to determine the optimal white balance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If face detection is used for white balance adjustment, then white balance correction can be applied to portrait images, but erroneous correction occurs when non-human faces or painted faces are detected
Solution Approach 1:
The system dynamically adjusts the face detection criteria and correction application based on the detected face characteristics. It evaluates whether the detected face area corresponds to a human face with skin color, and only applies face-specific white balance correction when the criteria are met, otherwise falls back to standard white balance correction.
Solution Approach 2:
The system incorporates feedback mechanisms by comparing the detected face area characteristics against expected human face properties (skin color ranges, facial features). This feedback loop determines whether to apply face-based or standard white balance correction, preventing erroneous corrections on non-human or painted faces.
2Measurement precision
If face area white balance correction is applied, then accurate correction is achieved for human faces, but the system cannot distinguish between human faces and non-human faces
Solution Approach 1:
The system applies different white balance correction strategies to different local regions or cases. For human faces with skin color, it applies face-specific correction; for non-human faces or painted faces, it uses standard correction. This local differentiation resolves the contradiction by tailoring the correction method to the specific face type detected.
Solution Approach 2:
The system changes the detection parameters and correction algorithms based on the detected face characteristics. It monitors parameters such as skin color presence and facial feature patterns, and switches between correction modes accordingly, enabling accurate discrimination without requiring complex additional detection mechanisms.
3Adaptability or versatility
If standard white balance correction is used, then general applicability is maintained, but accurate correction fails in portrait images with skin tone requirements
Solution Approach 1:
The system dynamically selects between standard and face-specific white balance correction methods based on the presence and characteristics of detected faces. When a valid human face is detected, it switches to the more precise face-specific correction; otherwise, it maintains the general standard correction, thus achieving both adaptability and precision.
Solution Approach 2:
The white balance correction system is designed to perform multiple functions: it can apply both standard white balance correction for general images and face-specific correction for portrait images. This multi-functionality allows it to maintain broad applicability while achieving high precision when needed.
Data Source
AI summary
A normal AWB (auto white balance) correction value is calculated based on inputted image data. Further, a face area is identified from the inputted image data and a face AWB correction value is calculated based on image data in the face area. Then, first feature data and second feature data are extracted from the inputted image data and image data in the face area, respectively. A total AWB correction value is calculated in accordance with at least one of the face AWB correction value and the normal AWB correction value based on a comparison result of the first feature data and the second feature data. Thus, an erroneous correction can be prevented in an AWB correction using a face detection function.