Abstract:

Face recognition is an important part of biometric identification. The eyes are one of the most prominent features of the face, and their localization is a key step of face recognition. Integral projection is usually used for eyes localization, but the horizontal position of eyebrows is always mistaken for eyes’when we use this method directly. This can lead to accuracy reduction. So after roughly marking out the area of face, we add the differential projection, namely, the gray value of the region that changes frequently in the horizontal direction, to the integral projection. The accurate position of eyes is defined by combining the integral projection and the differential projection. Experiments on the ORL face database show that the proposed method can get 90.5% accurate positioning performance, thus locating eyes more exactly.

Key words: eyes localization, integral projection, differential projection