Robust Face Tracking Using Self-Skin Color Segmentation

This paper addresses the problem of human face tracking in color image sequences. A method using human skin color feature integrated with condensation algorithm is proposed. At first, this paper presents a method of self-skin color segmentation to detect human skin in color image. Compared with conventional skin color model methods, this method has more precise detection performance in varying lighting condition. Additionally, this paper presents two improvements for condensation algorithm: a simple and efficient sampling method based on the Metropolis algorithm and an adaptive dynamical model are used for robust face tracking. Experimental results show that this approach could track human face's free movement, even in complex backgrounds and occlusion case