Performance Analysis of Gabor Responses in Face Recognition

In this paper, we investigate the behavior of Gabor responses a t automatically located facial feature points for face recognition. In our approach, a set of feature points on the facial landmarks is selected, where features of interest are eyes, nose, mouth and left and right contours of the face. As a preprocessing step all images in the database are normalized, eye positions in the normalized images are set a t preset fixed coordinates, and integral projection is used for automatically locating facial landmarks such as nose and mouth areas in face images. Finally, Gabor responses a t respective feature locations are extracted. The extracted features are analyzed for recognition performance using a neural network classifier with backpropagation learning, where input to the network is a similarity vector corresponding to feature points of two faces.