A simple and efficient eye detection method in color images

International audienceIn this paper we propose a simple and efficient eye detection method for face detection tasks in color images. The algorithm first detects face regions in the image using a skin color model in the normalized RGB color space. Then, eye candidates are extracted within these regions. Finally, using the anthrophological characteristics of human eyes, the pairs of eye regions are selected. The proposed method is simple and fast, since it needs no template matching step for face verification. It is robust because it can deals with face rotation. Experimental results show the validity of our approach, a correct eye detection rate of 98.4% is achieved using a subset of the AR face database