Face recognition in a large dataset using a hierarchical classifier

Face recognition is one of the most common authentication methods. Although much research has been conducted in this area, there are still many challenging issues to be addressed on face recognition, such as a large number of images in a dataset, with only one sample per person. The goal of this paper is to provide a robust face recognition method for a database having a large number of images with only one sample per person. The proposed method first uses a simple clustering approach to divide the images hierarchically into balanced clusters. Balanced clustering helps us to continue clustering in several hierarchies and finally reach very small clusters of equal size. Then, the face recognition task is performed within each cluster. A combination of the Non-negative Matrix Factorization (NMF) and the Fast Retina Key-point (FREAK) descriptors has been used to match the faces. The proposed method was evaluated on the FERET dataset that achieved an accuracy of 98.36%. Also, some other experiments have been done to validate the efficiency of the proposed method. The results of the experiments show that the proposed method can be applied to even larger datasets, while its complexity increases linearly.