A Two-Stage Color Correction Algorithm for Facial Image
Huilin Liu
Basic Medical College, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Lijuan Wang
Laboratory of TCM Four Processing, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Peng Qian
Basic Medical College, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Ying Xu
Laboratory of TCM Four Processing, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Zhumei Sun
Basic Medical College, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Chunrong Guo
Basic Medical College, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
Fufeng Li *
Laboratory of TCM Four Processing, Shanghai University of Traditional Chinese Medicine, 1200 Cailun Road, Shanghai, 201203, China.
*Author to whom correspondence should be addressed.
Abstract
Aims: Facial image analysis is the main direction of facial diagnosis objectification, where realistic colors in medical facial images are essential clues for disease diagnosis. However, the same person's facial color information will have different display information under different light sources and display devices, which may lead to different diagnostic results.
Methodology: Firstly, we conducted group experiments on the selection of color patches and selected 12 face-related color patches for color correction of facial images. Next, to adapt to this task under different lighting conditions, we utilize an adaptive white balance algorithm to adjust brightness as the first step. We selected the D65 illuminate environment as the standard lighting condition. Finally, we employed a polynomial-regression algorithm based on ridge regression for color correction.
Results: The experimental results demonstrate that the average values of the chromatic distance of our method under all five lighting conditions are less than 3.0. We also compare the ablation experiments of adjusting brightness.
Conclusion: The results indicate that the proposed method has good consistency in color correction of facial images under different lighting conditions, which is more suitable for clinical medical diagnosis.
Keywords: Color correction, facial image, white balance, regression model