Image Retrieval Using Efficient Region-based Matching

Abstract

Identifying regions of interest (ROI) plays a vital role for humans to find desired images in content- based image retrieval (CBIR). To enhance the performance of CBIR systems, we propose a simple but generally effective model to express ROI rather than pursuing sophisticated image analysis techniques. Our approach partitions an image into a number of regions with fixed absolute locations. User can formulate a query by selecting the interesting regions in the image. Candidate images are then analyzed, by inspecting each region in turn, to find the best matching region with the query region. Experimental results show that the presented model is generally effective and particularly suitable for images with regions having features which significantly differ from the global image features.