Alireza Rahmani Azar - Department of Computer Engineering, Faculty of Engineering Arak University Arak 38156-8-8349, Iran
Farhad Khalilzadeh - Department of Computer Science Hacettepe University Ankara, Turkey

Nowadays research in face features detection and recognition is getting more attention among researchers due to its different applications in science. This paper proposed a new method for eye detection using edge detection and Euclidean distance. In the proposed method, skin region is specified using a skin detection technique which employs some formulas in two different color spaces. In the specified skin region, horizontal edges are detected by combining the results of two known masks, Prewitt and Sobel. By using morphology operations such as dilatation and erosion, very small edges are removed and close edges are connected. The face region is divided to three parts: the upper right quarter, the upper left quarter and the lower half. In each part, the biggest binary object is found as the right eye, the left eye and the mouth. In order to verify the previous steps, the distances between left eye center to mouth center and right eye center to mouth center are found. The proposed method is tested on PICS, which is a database containing face images. It shows 93 percentage accuracy in an acceptable time.

eye detection; edge detection; human computer interface