Research on Multi - feature Web Image Clustering Algorithm

Abstract

In order to find interesting images from massive Web resources, mine useful information, the clustering algorithm of multi-feature Web images based on Web2.0 is studied in the paper. By using this algorithm, the relationship between the Web image and its metadata is structured a K-partite graph by a reasonable measure of metadata similarity. Clustering results can be obtained by the K-partite graph. The accuracy of clustering can be enhanced through the effective fusion of rich heterogeneous metadata in Web image. High-quality clustering results obtained by the algorithm can be used to mine useful information from web images. Users do not need to give the weight of each type of metadata. The algorithm researched in the paper is scalable, which can be applied to large-scale image clustering and can be parallelized.