Illumination-robust face recognition via sparse representation

Sparse representation receives a lot of attention as a representation method of signals in recent years. The sparse representation-based classification (SRC) applies sparse representation to face recognition. Camera parameters and/or illumination change may occur in real situations of face recognition. This study aims to propose a illumination robust SRC. The proposed method modifies the bases which are added to training bases derived from the training data set to detect noise contained in the recognized facial image. In the case where a few training data set is prepared, this method succeeds to obtain the higher recognition rate than the conventional SRC.