Face Recognition Using Improved Co-HOG Features

Face recognition is one of the most sought-after biometric technologies in the field of machine learning and computer vision in recent years. Histogram of oriented gradients (HOG) descriptor was originally developed for human detection and recently, it is also being applied to face recognition. However, when compared with other successful feature descriptors such as SIFT, LBP, Gabor, and so on, there is still considerable research space on the application of HOG features for face recognition. Co-HOG, a variant of HOG, uses a pair of gradient orientations as its basic building block, unlike HOG which uses a single gradient. Using the pair of orientations, the co-occurrence matrix is computed and histograms are calculated. However, in Co-HOG, gradient direction alone is considered and magnitude is ignored. It is believed that gradient magnitude also carries significant information about features. In this paper, we develop a face recognition system that utilizes Co-HOG features with embedded gradient magnitude information. The experimentation is done on ORL face dataset and it is observed that the proposed model is better than other existing methods with a maximum accuracy value of 97%.