A novel design for face recognition system using Dual Tree Complex Wavelet Transform

Abstract

Face recognition is the process of identifying a person by grabbing his face and matching it against a trained database. Facial feature extraction is one of the methods for face recognition. We intend to propose an efficient facial feature representation by using Dual Tree — Complex Wavelet Transform (DT-CWT). The Complex-WT face characterizes the geometrical structure of facial images by using the properties of DT-CWT such as approximate shift invariance and good directional selectivity. Since the efficiency retained with DT-CWT is inadequate, a new block design using Dual Tree Complex Wavelet Transform along with efficient normalization and noise reduction techniques is developed. Normalization is done using Discrete cosine transform and Total variation minimization approach is used as a noise reduction technique. Using the new design a face recognition system is developed. Effectiveness of the system is evaluated using FERET and Yale databases