Real time detection of eye corners and iris center from images acquired by usual camera

Abstract

A precise iris center and eye corners detection algorithm is proposed. The work is based on common web cameras and natural light condition instead of high resolution devices and infrared illumination. The algorithm includes two parts. Firstly, the face region is detected with Adaboost method. Moreover, Active Appearance Models is used to locate the rough area of the eyes. Secondly, an improved local projection function and circle integral algorithm is presented. Based on this algorithm, the iris center and eye corners in the eye region can be precisely located. The experiments show a high accuracy. Compared with the existing methods, this algorithm provides a more reliable application in the gaze based aids tools for disabled people, such as cerebral palsy patients.