Skin Pigmentation Detection and Measurement based on ICA Algorithm

Abstract

Skin color of human is combined by the color which is from the melanin, hemoglobin and carotene etc. It is different distribution in different areas and crowd. When our skin become damaged or unhealthy, it affects skin pigmentation production. skin pigmentation not only influence people's looks, some skin pigmentation disease will affect people's health. Skin pigmentation detection is very important in the skin disease diagnosis. Accurate detection is the premise of the fast cure skin disease. In evaluating the efficacy of medical hairdressing and cosmetic, skin pigmentation detection is also a important basis. This paper presents an approach for detecting and measuring the skin pigmentation of human based on independent component analysis (ICA) algorithm. In the proposed scheme, we extract the skin area by using Gaussian skin color model in YCbCr color space and remove some noises by the morphology processing. Next, we decompose the skin area into two components: hemoglobin and melanin, by ICA algorithm and calculate local and global intensities, locations, and sizes of the two components by the 2D location histogram. Then we determine the existence of skin pigmentation disorders by comparing local and global intensities of them. And we give two parameters pigmentation area percentage and image standard deviation for the skin pigmentation quantification. We tested our scheme with images in several different illumination conditions and also tested it on several cosmetic covering performances. As our expected, experimental results of our scheme has better illumination adaptability and processing time than the conventional scheme. And our scheme also can be used to measure the vision-based evaluation of cosmetic covers by the parameters