A comparison of human skin color detection for biometrie identification

Human skin color detection performs a critical part in the biometric identification and recognition system. In this paper, androgenic hair pattern recognition system compared 10 rules of human skin color detection, employing color spaces such as RGB, HSV, YCbCr, CIE Lab and YIQ color spaces. There were 400 images of androgenic hair pattern investigated, obtained from 25 male respondents. Detected human skin area images were converted into grayscale images and binary images. Precision of the recognition system was measured for each rule by employing 2-fold cross validation and Euclidean distance to calculate the nearest neighbor. The recognition system did not use any feature extraction algorithm. The experimental results showed that binary images were more suitable for simple recognition system without any feature extraction method and the best recognition precision was obtained by using color space that divided luminance component and chrominance component such as YCbCr, CIE Lab, HSV, and applied only chrominance components to the rules of human skin detection. The best precision recognition, achieved by utilizing YCBCr color space (using of Cb and Cr only), was 84.35 % from database with binary images.