Face Recognition Using Effective and Fast Approach

Dr. Ali Mohammed Sahan IT Dept., Technical College of Management, Middle Technical University, Baghdad, Iraq AbstractFace recognition process in human perception utilizes two kinds of features. The first are global features used to describe the whole face image, while the second are local features used to describe finer details of the face. The combined face recognition methods stimulate the technique used by human perception to recognize the face images in a natural way. Therefore, the combinations of global and local features have received significant success in face recognition problem. In this paper, we propose a combined global and local face recognition method. The global features are obtained using complex Zernike moments (CZMs), while the local features are captured using a modified form of the local binary pattern (LBP) descriptor. The CZMs use both the magnitude and phase of complex moments to represent global features effectively. We use fast algorithm to compute Zernike moments (ZMs) to increase the speed of the proposed method. A new form of LBP which uses a 5 5 window instead of the convention 3 3 window has been developed which enhances recognition rate significantly. Extensive experiments have been conducted on five different face databases, namely, FERET, JAFFE, YALE, UMIST, and ORL to evaluate the performance of the proposed method under different face recognition conditions. Results of these experiments reveal that the proposed method is much more robust against facial expression, pose, and illumination variations than the existing combined approaches. KeywordsFace recognition, Complex Zernike moments, Local binary pattern, Global features.