Eye location method based on Haar features

Eye location is the imported reference of face normalization in face recognition system. This paper describes a novel method to improve the eye location accuracy in the face image, which detected eyes by combining Haar features, dual-orientation Gabor filters and eye templates. First, introduce Haar features to detect face region from face images, and then detect the binocular region based on Haar features in the face region. Finally accurately locate eye coordinates by combining dual-orientation Gabor filters and eye templates from the preliminary binocular region or face region. The experimental results on images of Face Recognition Technology (FERET) database demonstrate that this proposed method was able to achieve a location accuracy of 86.34% in 1477*2 eyes, and it is effective in video images by experiments. Using this method can enhance the performance of face detection and recognition in the video surveillance system.