Location of eyes in face images

Abstract

This paper describes a method for detection of irises in face images. The approach follows the work described in [1] with new procedures introduced by the authors in order to improve the detection results. The global approach has the following main steps: image cropping; image "cleaning"; selection of possible eye locations; costs computation for each candidate predetected eye; determination of the "best" pair of "eyes". The innovation introduced is mainly in the first two phases: the image cropping process and the image cleaning. Completely new code was written in Matlab®, providing a user friendly graphical interface which paves the way for the making, testing and debugging of the algorithm. This work was developed within a Digital Image Processing course at the Electrical and Computer Engineering degree in the Faculty of Engineering, University of Porto.