An Eye-Contour Extraction Algorithm from Face Image usingDeformable Template Matching

A variety of studies on face components such as eyes, lips, noses, and teeth have been proceeding in medicine, psychology, biometrics authentication, and other areas. In this paper, we present an algorithm of extracting eye contours from a face image using the deformable template matching method. Our template for an eye contour is composed of three quadratic functions for the perimeter and one circle for the pupil. In our algorithm, a digital color face image is rst converted to a binary image of representing eyes, after the region around eyes is identied on the face image by using hues and values of the color image. Then, parameters in the template are optimized by a local search method with a tabu period and a hill-climbing, so as to t the template to the eye contour in the binary image. The accuracy of our algorithm is evaluated through sample face images of students. In addition, the application of our proposal to eye shape indices is investigated in a face image database "HOIP", where recognizable dierence exists in index distributions between males and females.