Multi-race face detection based on skin color and eyes

A multi-race face detection system based on skin color and eyes is presented for images. It consists of skin color segmentation, template matching and eye location. In skin color segmentation, an additional channel for dark color race is added based on different skin color clustering characteristic between dark and light color race in color spaces. This additional channel can increase the number of face region candidates for dark color race. Consequently, template matching and eye location is to decide whether a face exists in these region candidates. The experimental results show that the proposed method is effective for multi-race faces.