Study of image retrieval based on user defined region of interest

The image retrieval system based on the global character can't meet the needs of users and that based on the image segmentation excessively depends on segmentation algorithm. To overcome shortcomings of these two systems, according to region-based image retrieval, a new image retrieval algorithm based on user defined region of interest(ROI) was presented. First, ROI was designated and partitioned into unequal-sized sub-images by using the image segmentation technology. Then, the main color of each block was extracted and the main color histogram was computed as the image's color feature. The sub-images were quantified into different types by using flatness and roughness. Histograms of different blocks were adopted as the image shape feature. The tests show that the system is more effective in image retrieval.