Using SVM to Organize the Image Database

In a content-based image retrieval (CBIR) system, rational and effective organization of the image database plays an important role in improving the performance of the system. In this paper, we propose a new method to classify the images database of CBIR system. Using SVM we attempt to construct a mapping between the low-level features and the semantically level in order to determine which category an image belongs to. The selection of training set is different from human determined and we consider use affinity propagation (AP) clustering method to generate them. Different numbers of clustering can complete classification which satisfied different retrieval need of user, such as exact match or rough match. In the experiment, we choose flower, forest and sky as experimental images. The accuracy of exact classification and rough classification is satisfactory.