3D transformation based feature extraction and selection for 3D face recognition

This paper presents a novel method for 3D facial feature extraction. In this method, 3D discrete cosine transform (DCT) is used to extract features. Before the feature extraction, faces are aligned with respect to nose tip and then registered two times: according to average nose and average face. Then the coefficients of 3D transformation are calculated. The most discriminant 3D transform coefficients are selected as feature vector where the ratio of within-class variance and between-class variance is used as selection measure. To further improve the class discrimination, linear discriminant analysis (LDA) is applied to feature vectors before the matching procedure. We also investigate the proposed approach by using 3D discrete Fourier transform instead of 3D DCT for comparison. Selected and discriminated features are matched by using the nearest neighbor classifier. Recognition experiments are realized on 3D RMA face database. The proposed method shows the recognition rate above 99% for 3D DCT based features.