Automatic feature extraction of facial organs and contour

This paper deals with algorithms for extracting contour lines of eyebrow, eye, nose, mouth, and face contour. First we extract the position of both irises using a correlation technique. Then for the eyebrow, edge-image data are used for determining the position of the eyebrow and the contour of the eyebrow is obtained from binary image data around the eye area. For eyes, the edge between the eyeball and skin is determined easily since the edge is clearly differentiated in the brightness distribution of the image data. For the nose, the edge clearly appears in the image data and then the correlation can determine the nose contour. For the mouth, the position of mouth is determined from image-edge information and brightness distribution and then the correlation can determine the most outer corner points of the mouth. We employ an image energy function to be minimized for obtaining the face contour and a dynamic programming like algorithm is used to obtain the minimum. An average human face contour used as a template can give a good extraction of the face contour.