Improving CBIR systems using clusterization

In this paper Content Based Image Retrieval (CBIR) system with user assistance for searching clustered image databases is presented. In the preprocessing phase a image database is clustered on class level, and represented with small number of images. By implementation of this system the initial search of the whole system is significantly reduced, and the first iteration of user assistance is automated. System is tested on Corel 1K image database and compared with the standard realization of the CBIR system with user assistance. The experiments are shown reduced searching time and high searching efficiency of presented system.