摘要

This paper proposes an automatic eye feature extraction method based on Hough transform and eyelid deformable template. It could locate the iris center and the contour of the eye automatically and correctly. Compared with other similar deformable template methods, this approach has the following advantages: It could be used to process those low contrast images in which the lower eyelid is blur. It could extract accurate feature. It simplifies the difficulty in selecting empirical coefficients by this improved energy functions and conservative energy descent algorithm (CEDA). Its efficiency and superiority are clarified by experimenting on about one hundred eye images.