基于显著性和视觉词典的图像分类算法研究

In this paper，an image classification algorithm is proposed to deal with the blind image classification, which means the content of image is unknown, but the classification task could be guided by human visual perception. Firstly, the salient regions will be extracted from natural images; then compute the SURF features of the salient region; thirdly build up the visual vocabulary of the salient region; the classification could be worked out by 1-norm SVM' s sparse property. The experimental results show that the efficiency and effectiveness of the presented approach.