Nostrils and Chin Segmentation

This report describes our approach to the localization of the position of nostrils and chin. Basic demand is a good functionality on pictures of face without special make-up. The input picture contains face, hair and shoulders of observed person. We don't use any special lighting conditions for capturing the input images, therefore our program is useful in many common cases. Mentioned algorithm will be used in the initial steps of very complex process of Automated Recognition of Facial Expressions (ARFE). Good start is a basic condition for success of ARFE, then algorithm have to be steady. Nostrils are used for face coordinate system generating and chin makes lower limit of the face area. In the rst section of our report we introduce the basic problems of ARFE and we describe main purpose of our work. The second section deals with detail input and output speciications and restrictions. In third part we explain diierent approaches of solution with their diicul-ties. Next section is devoted implementation of our algorithm with detail description of particular steps. In the nal parts of this paper we introduce realized experiments and think about usage and improvement of our program.