In this paper, we present a novel approachfor face recognition which consists of a dimensionalityreduction of face feature vectors. The image scaling is firstlyconducted on an input face image. Then we applied the LocalBinary Pattern (LBP) operator by dividing the face imageinto non-overlapped regions. LBP histograms are extractedfrom each region and concatenated into a single one thatrepresents the face image. Nearest neighbor classifier isused to perform recognition with Chi square function asa dissimilarity measure. Simulation experiments are doneusing the ORL (Olivetti Research Laboratory) databaseshowing the efficiency of the proposed approach with 97.5%as recognition rate.