Significant region-based image retrieval

In the era of multimedia technologies, the need for information/data retrieval systems getting more attention. The data might be image, video, audio and/or text files. Digital libraries, surveillance application, web applications and many other applications that handle huge volume of data essentially have data retrieval components. In this paper, a technique of region-based image retrieval (RBIR), a branch of content-based image retrieval (CBIR), is proposed. The proposed model identifies a significant region in an image using visual attention-based mechanism and represents them using its color layout descriptors and curvelet descriptors. These features are extracted from significant region of query image and images in the database. The likeness between the query region and database image region is ranked according to a similarity measure computed from the feature vectors. The proposed model does not need a full semantic understanding of image content, uses visual metrics such as proximity, size, color contrast and nearness to image's boundaries to locate viewer's attention and uses curvelet transform in combination with color layout descriptor to represent the significant region. Experimental results are analyzed and compared with the state-of-the-art CBIR Systems and RBIR Technique.