Mining of Images by K-Medoid Clustering Using Content Based Descriptors

Image Mining is a challenging task in the data mining and image processing field. In Image mining, useful information is extracted from the enormous collection of image database. The expansion of images has been risen up in each areas medical field, business, forgery detection etc because images easily gain the attention to the users that's why more researchers are attracted towards this field. Lots of images are scattered in the database so to manage the database, clustering is applied which is one of the techniques of Image Mining. In this research work fusion of color and shape features are used for extracting the descriptors from the images through Color Moment (CM) and Edge histogram Descriptor (EDH). After that K-medoids Clustering algorithm is applied on the created dataset to obtain clusters. Finally, the output of clustered images will be shown. Three databases are used for testing this system Wang (1000), Coral (2000), Oliva (452). By using Precision, Recall, F-measure and Error rate metrics, we measure the performance of this proposed work and will also compare with other's conventional methods.