Face Recognition Using the Improved Bag of Words Model

Bag of words (BoW) model, which was originally used for document processing field, has been introduced to computer vision field recently and used in object recognition successfully. However, in face recognition, the order less collection of local patches in BoW model cannot provide strong distinctive information since the objects (face images) belong to the same category. A new framework for extracting facial features based on BoW model is proposed in this paper, which can maintain holistic spatial information. Experimental results show that the improved method can obtain better face recognition performance on face images of AR database with extreme expressions, variant illuminations, and partial occlusions.