Face recognition system using discrete wavelet transform and fast PCA

Abstract

The face recognition system is used to create a national database for the purpose of identity cards, voting in an electoral systems, bank transaction, food distribution system, control over secured areas etc. In this paper we propose the Face Recognition System using Discrete Wavelet Transform and Fast PCA (FRDF). The Discrete Wavelet Transform is applied on face images of Libor Spacek database and only LL subband is considered. Fast Principal Component Analysis using Gram-Schmidt orthogonalization process is applied to generate coefficient vectors. The Euclidean Distance between test and database face image coefficient vectors are computed for face recognition based on the threshold value. It is observed that the face recognition rate is 100% and the proposed algorithm for the computation of eigenvalues and eigenvectors improves the computational efficiency as compared to Principal Component Analysis (PCA) with same Mean Square Error (MSE). © 2011 Springer-Verlag.