No abstract

With the explosive growth of Web development of digital image processing t applications of Web image have attracted much the Web image retrieval. Since the Web imag some related text tags, making use of both v features of Web image will help improving the Web image clustering. Researches show t clustering methods, such as graph partition hypergraph partitioning models, didn't make between texts and image simultaneously. In explore to take both visual and textual features Web image clustering by building a graph mod novel iterative clustering method. With K clust calculate the occurrence frequency of each visu over the j-th cluster (j = 1, 2, …, K), which is use significance of the feature for the j-th cluster. Th of each image, which belongs to the j-th determined accordingly. Furthermore, a mixtu for the predicted feature linked to each ima algorithm is adopted to get K component p describe posterior probabilities of all clusters Then two K-dimensional vectors consisting parameters will be used to describe the imag cluster index of it. Several experiments have bee MIR-Flickr25K and IAPR TC-12 Benchmark performance of the proposed Web image cluste superior to that of the compared algorithm.

No abstract