Fast Computation of Mutual Information in a Variational Image Registration Approach

Stefan Heldmann, Oliver Mahnke, Daniel Potts, Jan Modersitzki, Bernd Fischer


In this paper we present a novel method for computing the Mutual Information of images using recently developed non-equidistant fast Fourier-transform (NFFT) techniques. Standard approaches suffer from the problem that some sort of quantization is needed in order to apply a fast equidistant FFT-technique. For the new method no quantization is necessary which on one hand leads to an improved registration accuracy and on the other hand to a straight forward implementation. The evaluation is done for MR brain registration as well as for synthetic examples.
Original languageEnglish
Title of host publicationBildverarbeitung für die Medizin 2004
EditorsThomas Tolxdorff, Jürgen Braun, Heinz Handels, Alexander Horsch, Hans-Peter Meinzer
Number of pages5
Place of PublicationBerlin, Heidelberg
PublisherSpringer Berlin Heidelberg
Publication date01.03.2004
ISBN (Print)978-3-540-21059-7
ISBN (Electronic)978-3-642-18536-6
Publication statusPublished - 01.03.2004
EventWorkshop on Bildverarbeitung fur die Medizin 2004 - Berlin, Germany
Duration: 29.03.200430.03.2004


Dive into the research topics of 'Fast Computation of Mutual Information in a Variational Image Registration Approach'. Together they form a unique fingerprint.

Cite this