Eye detection based on the Viola-Jones method and corners points

Abstract

Eyes detection is a very interesting field of research that verifies the presence of eyes and locates their positions in an image. Similarly, it is often the first step in such applications such as face recognition, human machine interaction systems, facial expression recognition, and driver fatigue monitoring systems. In this paper, we proposed a robust eye detection method based on the Viola and Jones method and corner points. Firstly, faces are detected by a system composed of two detectors of Viola-Jones (one for the frontal faces and the other for the profile faces). Secondly, we used the Shi-Tomasi detector (to detect corner points) and K-means (for clustering the neighbor corner points) to determine eye candidate regions. Thirdly, the localization of eyes is achieved by matching of these regions with an eye template. The results obtained show that our method is robust and provides superior performance compared to other recently published methods.