Image clustering based on SIFT-affinity propagation
Yanpeng Zhang, Hong Zhang · 2014
Clustering is a hotspot issues in the field of data mining. There is abundant digital image information in the image acquisition equipment, the image database or the Internet. Facing the large scale image information with rich semantics, it is difficult to obtain accurate information as soon as possible. Therefore, it is essential for us to study efficient image clustering algorithms, in which how to measure the correlation among the samples is key issue, and will directly affect the results of clustering. In this paper, a new image clustering function based on SIFT-Affinity propagation is proposed. First, we extract image features based on Scale-invariant feature transform(SIFT), which are changeless to image zooming, translation, and gyration, and partially changeless to brightness changes and affine. Then the modelling of correlation is established and quantitatively analyzed. Finally, we utilize affinity propagation (AP) to fuse different SIFT correlations among images and obtain clustering results. Our method considers all image data as likely clustering center. The examples results prove the effectiveness and superiority of our approach.