Face Retrieval Using 1st- and 2nd-order PCA Mixture Model

Abstract

This paper deals with face retrieval using the 1st- and 2nd-order PCA mixture model. The well-known eigenface method uses one set of holistic facial features obtained by PCA. However, the single set of eigenfaces is not enough to represent the complicated face images with large variations of poses and/or illuminations. To overcome this weakness, we propose the (1st-order) PCA mixture method that uses several eigenface sets obtained from the EM learning in PCA mixture model. Furthermore, we propose the 2nd-order PCA mixture method that combines the 2nd-order eigenface method and the PCA mixture model. In the 2nd-order eigenface method, each image is represented by a couple of feature vectors obtained by projecting the face image onto a selected approximate eigenface set and then by projecting the residual face image onto a selected residual eigenface set. Face retrieval is performed by finding the identity that provides the shortest distance in the feature space between the input image and the template image. Simulation results show that face retrieval using the 2nd-order PCA mixture method is best for face images with illumination variations and face retrieval using the 1st-order PCA mixture model is best for the face images with pose variations in terms of ANMRR (Average of the Normalized Modified Retrieval Rank) and CRR (Correct Retrieval Rate).