Chest radiographs retrieval based on texture, shape and semantic information

This paper casts about for effective retrieving technique from imagepsilas texture, shape and semantic information, in allusion to chest radiographs, the Gray level co-occurrence matrix with Gray-Grads co-occurrence matrix in texture feature and Hu moments with Zernike moments in shape feature are compared. A prototype system is implemented. With experiment, we find the Gray-Grads co-occurrence matrix getting better retrieval performance than the Gray level co-occurrence matrix in texture, the Zernike moments getting better retrieval performance than Hu moments in shape, and an effective technique making use of texture, shape and semantic information to retrieve which companied with relevance feedback has been used. The comparison of the retrieval results show that the approach proposed in the paper is effective.