Illumination Invariant Measuring of Skin Pigmentation

This paper proposes a new method for measuring concentration of skin pigmentation using nothing but a standard RGB camera. Since RGB signals are easily affected by change in illumination, all conventional methods require a prior knowledge about the illumination or some constrains about photographing device. The proposed method obtains robustness against change in illumination by calculating ratio between signals of skin regions with and without pigmentation. Since illumination within local area is considered to be uniform, the ratio calculated from these two signals within local area is transformed to a linear combination of following three values. The values represent attenuation of light through skin pigmentations; melanin and hemoglobin, and attenuation of light associated with asperity of the skin surface, respectively. We derive coefficients of the each term by minimizing the least square error to obtain concentration of skin pigmentation. In order to evaluate accuracy of the proposed method, we compute amounts of pigmentation under eight light sources without the prior knowledge and any constrains. This result shows that for most people the average error is small enough not to recognize. It shows the proposed method is robust against change in illumination. Our method enables people to examine their pigmentations under any kind of illumination even in their homes, without visiting cosmetics counters. The one of other effective applications is to remove only skin pigmentation on a photograph without blurring skin texture. This enables to improve face image quality without losing the high resolution characteristics of the displays.