Mouth and Nostril Detection based on Facial Skin Color Distribution

Bo Peng, Nobuaki Nakazawa, Motohiro Kano

In this paper, we proposed a new detection method for a face feature point, based on facial skin color,independent of tilted face angles.We picked up the nostril and mouth as a face feature point. Human’s face was observed by a camera in real time and the skin color was utilized to detect the face domain. In order to detect the position of nostrils, we focused on the fact that the brightly-colored facial area is distributed around the nostril while the dark-colored facial area is distributed around the eye, the mouth and the jaw.Here we estimated the possibility area of existing nostril by calculating of the red component ratio. Within the estimated area, we detected the nostril position based on the conditionsa) The size of the component is close, b)    The inclination of the straight line passing through two points is small, c)    Short distance between two points. Furthermore, according to the relative position between the mouth and nostril, the mouth position could be detected.  In the test trial, we confirmed that the developed system could detected the nostril and mouth of subject successfully.