A NEW SIMILARITY METRIC BASED ON 2D VECTOR FOR ICONIC IMAGE RETRIEVAL

Abstract

Recently, the requirements of multimedia information are increasing rapidly. The efficient similarity images retrieval from a large image data bank being the purpose of this article. The spatial relationship is an important feature in content-based retrieval. In the studying of spatial relationship, previous related works all focus on the directional relationship, such as 2D-String. 2D C-String and 2D-PIR. etc. In this paper, we show that the distance is an important factor of spatial relationship. We propose a new representation to embed the distant relationship among objects in an image and a new similarity retrieval method. The Rnorm measure be used to evaluate the quality of our new method. and we find that the new simiarity retrieval method provides efficient performance and supports flexibility for users request.