A New Adaptive Color Space YCH for Face Detection

Face detection problem has been thoroughly studied in the computer vision community due to its fundamental challenges and interesting applications. Using information of skin color to detect face regions is a quick and effective way. However, it is very difficult to choose a suitable color space when designing a specific face detection algorithm since the discriminating power of the mostly used color spaces is relatively weak and lack of universality. Therefore, a novel adaptive color space YCH is first proposed in this paper. It fuses skillfully and seamlessly the merits of the commonly used YCrCb, rgb and HSV color spaces as well as the color ratio into a new simple nonlinear transformation constructed as one luminance and two chrominance components. All the transformation coefficients can adjust automatically according to the respective characteristics of each pixel in the face image and then eliminate effectively various unfavorable influence factors to the classifying result of the skin color and non-skin color. Experimental results demonstrate the good discriminating power of the proposed color space to all kinds of face images.