Discriminant Dual-Tree Complex Wavelet Features for Face Recognition

Abstract

In the field of face recognition (FR), the techniques that can provide effective feature representation with enhanced discriminability are crucial. Although the separable wavelet transform has played an important role in image processing, it suffers from some shortcomings which badly affect its ability for feature generation. In this paper, we propose a novel FR system combining the dual-tree complex wavelet transform (WT) and a two-step discriminant analysis, which is called discriminant complex- WT-face system. We introduce an effective feature construction method to provide more representative feature. Then the two-step discriminant analysis is used to filter and extract enhanced-discriminability features. Experimental results show that the discriminant complex-WT-face system can significantly outperform some other FR systems.