Joint spatial-spectral indexing of medical images exploiting the JPEG compression scheme

摘要

This paper approaches the problem of efficient content-based image retrieval, based on the JPEG compression standard. The proposed indexing technique takes advantage of this compression scheme by extracting frequency information from the DCT coefficients histograms, and spatial information from the gravity centers and inertia moments associated to the frequency histograms. Retrieval results are better than other techniques based on JPEG, or comparable to other compression schemes, but with a reduced complexity.