Face recognition using holistic Fourier invariant features

摘要

This paper presents a new method for holistic face representation, called spectroface. Spectroface representation combines the wavelet transform and the Fourier transform. We have shown that by decomposing a face image using wavelet transform, the low-frequency face image is less sensitive to the facial expression variations. This paper also proves that the spectroface representation is invariant to translation, scale and on-the-plane rotation. To handle the rotation in depth, multiple view images are used to determine the reference image representation. Based on the spectroface representation, a face recognition system is designed and developed. Yale and Olivetti face databases are selected to evaluate the proposed system. These two databases contain 55 persons with 565 face images at different orientations, scale, facial expressions, small occlusions and different illuminations. The recognition accuracy is over 94%. If we consider the top three matches, the accuracy is over 98%. The recognition system is developed on Pentium 200 MHz computer and the recognition time is less than 3 seconds for database with 55 persons