A boosting frame work for improved content based image retrieval

Abstract

This paper deals with medical image retrieval for retrieving images similar to query images from a database. Retrieval of archived digital medical images is always a challenge that is still being researched all the more so as such images are of paramount importance in patient diagnosis, therapy, surgical planning, medical reference, and medical training. This paper proposes using the Discrete Sine Transform (DST) for relevant feature extraction, and applies Boosting classification techniques to locate the relevant images. In this study, the boosting is used with J48 and decision stump. Experimental results show that the classification accuracy achieved is fairly good.