A method for content-based similarity retrieval of images using two dimensional DP matching algorithm

Abstract

A two dimensional dynamic programming (DP) matching algorithm is presented which is extended from conventional DP matching. In order to represent the color content of color images, a new concept, dominant color matrices is proposed. The retrieving strategy is divided into two phases to improve both speed and accuracy of the proposed retrieval. In the first phase, a group of candidate images, which are similar to the sample image, are produced by using the global color histogram intersection. In the second, dissimilar images in the group of candidate images are deleted by using the two dimensional DP matching algorithm. In addition, the proposed color-content similarity retrieval of images is used on an experimental database consisting of 120 varied color images. Experimental results show the proposed method is effective.