Parallel local histogram comparison hardware architecture for content-based image retrieval

The explosive growth of the World Wide Web has made a vast amount of information available especially the multimedia data such as images and graphics. New generation search engines with the technology of Content-Based Image Retrieval (CBIR) is a response to the need. In this paper, a new technique based on composite local colour histograms suitable for CBIR is described. The proposed approach is accurate even in very large image databases. Furthermore, a novel parallel hardware structure has been designed and implemented on a chip-set of FPGAs, in order to increase the operation speed. Its typical maximum clock frequency is 35 MHz and it can perform over 50 comparisons of 640 x 480-pixel images per second.