Abstract:

This paper presents a new face normalization method based on eyes location. First, the face is roughly located from the original image using a rapid face detection scheme hased on a hoosted cascade of simple Haar-like features. Second, we accurately locate the pupils of eyes in the face image according to the proportional relationships of face features and gray information. Then a coordinate system is established. Finally, we normalize the rotation, scale and grayscale of the face image. The experimental results show that this algorithm can detect and normalize the face image efficiently and accurately, and it can also improve the face recognition obviously.