A research on SRC plus homotopy in disguised face recognition

Abstract

Disguised face recognition is a great challenge to general face recognition systems, since a variety of disguises, to some degree, corrupt the information needed in identification. This chapter presents sparse representation-based classification (SRC) combined with homotopy algorithm to cope with disguised face recognition. To represent the face image sparsely, SRC constructs the overcomplete dictionary using training samples as atoms, and we employ homotopy to compute the expansion coefficients effectively. Experimental results based on Aleix Martinez and Robert Benavente (AR) face database show the validity of SRC combined with homotopy algorithm in face recognition. © 2012 Springer Science+Business Media B.V.