A new GLBSIF descriptor for face recognition in the uncontrolled environments

In uncontrolled environments, the major challenges in face recognition, such as illumination variation, occlusion, facial expressions and poses, greatly affect the performance of Facial Recognition Systems (FRS) especially those based on 2D information. We introduce, in this paper, a novel feature extraction approach named GLBSIF for face recognition in an uncontrolled environment. In our method, Gabor Wavelets (GW), Local Binary Patterns (LBP) and Binarized Statistical Image Features (BSIF) were combined. Moreover, the dimension reduction was applied in order to minimize the pattern vectors using PCA. Finally, we used KNN-SRC for classification. The introduced technique was assessed on LFW database using several experiments and tested on other databases, such as PUBFIG83, FERET, EXT.YALE B, ORL and IFD, in order to validate our approach. The best finding was provided when Recognition Rate (RR) is equal to 97.81%.