Eigenphases for Corrupted Face Images Naser Zaeri Electrical Engineering Dept., Kuwait University, P.O. Box 5969, Safat 13060, Kuwait zaery@eng.kuniv.edu.kw Abstract—Corrupted face image is one of the important obstacles that machine vision systems encounter when trying to recognize faces. In this paper, we propose a new face recognition system that can deal with the problem of corrupted images more efficiently. The new technique applies the principal component analysis to the phase spectrum of the Fourier transform of the covariance matrix constructed from the MPEG-7 Fourier Feature Descriptor vectors of the images. It will be shown that the proposed technique increases the face recognition rate when applied to images of low resolution and corrupted by noise, compared to other known methods. I. INTRODUCTION In the last few years, researchers in the area of face recognition have proposed many numerous techniques that achieve high recognition rate [1, 2, 3]. Despite the significant advances in face recognition approaches, it has yet to achieve levels of practicality required for many commercial and industrial applications. With the increased commercial interest in portable devices and with the advanced real-time face recognition systems, the need for more practical and cost effective systems has increased. For example, in surveillance systems face recognition plays an important role for reliable security issues. In such systems, the acquired images may be challenging as they suffer from noise and being of bad quality. Further, most of these images look corrupted and not in an acceptable shape for most of the existing machine vision systems. In this paper, we describe a new approach for face recognition system that can be implemented on such bad quality and destroyed images. It will be shown that the new technique provides improvements of the performance of the face recognition rates when compared to other approaches such as the fisherface LDA method, the direct eigenphase implementation method to the face space [4], and the PCA method. This work is organized as follows. Principal component analysis is described in brief in section 2, as it is used as part of our scheme. Another brief description of MPEG-7 Fourier Feature descriptor is given in Section 3. The formulation of the proposed technique is presented and discussed in Section 4. In Section 5, results of testing and implementing the new technique on the xm2vts database are presented. Concluding remarks are given in Section 6. II. PRINCIPAL COMPONENT ANALYSIS The PCA technique, proposed by Turk and Pentland [1], extracts the relevant information in a face image, encodes it as efficiently as possible, captures the variation in a collection of face images, and compares one face encoding with a database of models encoded similarly. The images of faces, being similar in overall configuration, will not be randomly distributed in the huge image space and, consequently, they can be described by a relatively low dimensional subspace. This process is achieved by finding the principal components of the distribution of faces, or the eigenvectors of the covariance matrix of the set of face images. The eigenvectors are ordered, each one accounting for a different amount of the variation among the face images. These eigenvectors can be thought of as a set of features that together characterize the variation between face images. Each image contributes more or less to each eigenvector, so that the eigenvector is displayed as a sort of ghostly face which is called an eigenface. Example of some of the images used from the xm2vts database and some of the resultant eigenfaces using the PCA is shown in Figures 1 and 2, respectively. Fig. 1. Examples from the xm2vts database. ACTEA 2009 July 15-17, 2009 Zouk Mosbeh, Lebanon 978-1-4244-3834-1/09/$25.00 © 2009 IEEE 537