摘要

A coarse-to-fine approach was presented. which can locate automatically an important feature often used in face recognition, the centre of the iris on human face. It combines such techniques as local binarization, Hough transform, pixels cluster and edge extraction. By this method, one could attain high localization accuracy whether the photo is taken under uniform or nonuniform light condition, whether the person wears glasses or not. So it is suitable for processing identification photos.