Abstract

In this paper a new feature vector, Wavelet Pyramid Based Binary Patterns (WPBP), is evaluated for Fingerprint Liveness Detection (FLD). It consists of two components: the first component involves detection of key points from four levels of pseudo-Laplacian pyramid obtained using Discrete Wavelet Transform (DWT) and their description using Local Binary Patterns (LBP) to represent multi-scale texture features; the second component consists of detection of shape, size and intensity variant features from first level wavelet approximation band. The features are then represented using Completed Local Binary Pattern (CLBP) descriptor. The combined feature vector is classified using Radial Basis Function (RBF) kernel Support Vector Machine (SVM) classifier. The proposed feature vector has been investigated for FLD on LivDet 2009, 2011, 2013 and 2015 competition databases. Experimental results demonstrate that the proposed feature vector is effective for FLD. The proposed feature vector is of reduced dimension, easy to implement and has good discrimination capability.