An Efficient and Robust Face Recognition System Based on Two-Dimensional Fisher Linear Discriminant Analysis:A New Approach

Eye portion of humans contains most discriminative information than any other part of the face [9]. This paper presents a new scheme of face image feature extraction. We made a successful attempt in obtaining good recognition rate by using only eye regions of a facial image. Initially, eye regions are segmented from the facial images. Segmented eye regions are then trained using two-dimensional fisher linear discriminant analysis (2DFLD) scheme. Based on experimentation it is observed that the proposed method requires fewer coefficients to represent an image and remained. insensitive to noise as compared to existing methods. The proposed method has been tested on AT&T database and compared with existing methodologies. The obtained results are described qualitatively and discussed.