Content-based retrieval of compressed medical images

Content-based retrieval of medical images has been shown to be useful for various applications. However, as images are typically stored in compressed form, a certain overhead in terms of computational cost, required bandwith and storage resources is associated with the retrieval process. In this paper we show that by performing retrieval directly in the compressed domain of the images this can be avoided. We employ a wavelet-based compression approach and make direct use of the derived wavelet coefficients for the calculation of image similarity and hence retrieval. Experiments performed on a database of 400+ medical infrared images show that our approach supports good retrieval performance.