Iris feature extraction using principally rotated complex wavelet filters (PR-CWF)

Abstract

Deriving effective iris feature from the segmented iris image is a crucial step in iris recognition system. In this paper we propose a new iris feature extraction method based on the Principal Texture Pattern (PTP) and dual tree complex wavelet transform (DT-CWT). We compute the principal direction (PD) of the iris texture using principal component analysis (PCA) and obtain the angle θ of the PD. Then, complex wavelet filters CWFs are constructed and rotated in the direction θ of the PD and also in the opposite direction - θ for decomposition of the image into 12 sub-bands using DT-CWT. Rotational invariant and scale invariant features are obtained by combining LL and HL sub-bands at each level. Consequently, channel energies and standard deviations are constructed as feature representation of the iris while SVM is used for classification of iris images. Our experiments demonstrate the superiority of the proposed method on CASIA iris databases, over existing methods.