Image Classification Method Based on Visual Saliency and Bag of Words Model

Based on traditional bag of words model and combined with the features of human visual, an improved image classification method is proposed in this paper. Firstly, compute the visual saliency of image. Secondly, according to the image visual saliency, we compute the histogram of visual words of image, and then use the histogram of visual words to represent the image. The validity of this method are carried out on Caltech 101 database. The experiment results show that the improved method performs better than other traditional method.