Application of Mutual Information in Image Retrieval
Baogen Xu · Dianzi Ke-ji Daxue xuebao · 2007
In this paper, a new content-based image retrieval method is proposed. By using the information entropy theory, the method describes the color, shape, and texture features of images, respectively. An integrated similarity match algorithm is presented. Firstly, an image is segmented into several sub-areas, and the angle information of the shape is computed after shape extracted from the image. Secondly, mutual information of color, shape, and texture among images is computed using information entropy. Compared with the other algorithms, experiments indicate that this approach has the advantages of good effect.