Face Recognition Using Histogram Oriented Gradients

Face Recognition Systems has been applied in a wide range of applications. However, their efficiency drastically diminish when they are applied under uncontrolled environments such as illumination change conditions, face position and expressions changes. Because of that, it is necessary to evaluate the performance of different feature extraction techniques robust to this kind of transformations for its further integration to a Face Recognition System. In this paper, we study and evaluate the pertinence of using the Histogram Oriented Gradients (HOG) method as a feature extraction technique to deal with the transformations already mentioned. To measure the performance of the proposed feature extraction method, several experiments were performed using two databases: one database under a controlled environment taken from the literature and other built in our laboratory under a semi controlled environment. The experimental results show that using HOG combined with different distances classifiers provides better results than those achieved with the well-know Eigenfaces technique.