Significant region based image retrieval using curvelet transform

Abstract

Region-based image retrieval system has been an active research topic in areas such as, entertainment, education, multimedia, image classification and searching. The system decomposes an image into discrete regions and each region is described using primitive features such as color, texture, shape or the combination of them. The extracted regions are indexed and retrieved. One of the key issues with the region-based image retrieval system is to extract essential information from the raw data which reflect the image content. Although large numbers of feature extraction, indexing and retrieval techniques have been developed, there are still no universally accepted techniques available for region/object representation and retrieval. In this paper we analyzes a biological vision based system which doesn't need full semantic understanding of image content, extracts features from significant/salient regions and index them for retrieval. The proposed system uses saliency map to locate viewer's attention and Curvelet Transform in combination with color histogram to represent the significant regions. Experimental results show that the proposed system outperforms the conventional image retrieval systems.