Multimodal Non-Rigid Registration


Medical image registration is a process in which two existing image representations of anatomy (e.g. MRI, CT, x-ray) or its function (e.g. PET, SPECT, fMRI) are put into correspondence. Registration approaches differ according to expected and modeled deformations and can be divided into rigid and nonrigid registrations. We are interested in nonrigid registration, which can detect and correct discrepancies of small spatial extent, by deforming one of the images (source) to match the other (reference). Spatial deformation model can be based on different physical properties like elasticity or viscosity, or their generalizations and simplifications. Deformation is driven by external forces, which tend to minimize image differences, measured by image similarity measures.
When images are acquired by different imaging procedures, we are talking about multimodal image registration. The problem of multimodal registration is in complex relation between intensities of both images, which is furthermore usually not known in advance. Measuring of image similarity is in that case based on statistics. The problem is even more difficult if images are supposed to be registered nonrigidly. To detect and correct small local image differences, external forces must be estimated from similarities of small image regions, but statistical significance of such small image regions is low and similarities are not sufficiently reliable and accurate.
We have focused on multimodal elastic matching to solve the problems mentioned earlier. We have introduced a new approach for measuring multimodal similarities of small image regions "point similarity measures". We have also built a system for multimodal nonrigid registration that is based on these measures. It uses a spatial model that is inspired by linear elasticity. Furthermore it uses multiresolution approach that improves registration speed and accuracy.


  • Peter Rogelj, Stanislav Kovačič. "Symmetric Image Registration". Medical Image Analysis, 2005.
  • Peter Rogelj, Stanislav Kovačič."Spatial deformation models for non-rigid image registration". In: Danijel Skočaj (ed.), Proceedings of the 9th Computer Vision Winter Workshop, CVWW'04 : 4-6 February 2004, Piran, Slovenia, pp.79-88. Slovenian Pattern Recognition Society, February 2004.
  • Peter Rogelj, Stanislav Kovačič, James C. Gee. "Point similarity measures for non-rigid registration of multi-modal data". Computer Vision and Image Understanding, 92(1): 112-140, October 2003.
  • Peter Rogelj, Stanislav Kovačič."Point similarity measure based on mutual information". In: James C. Gee, J. B. Antoine Maintz, Michael W. Vannier (eds.), Biomedical Image Registration : revised papers, (Lecture notes in computer science, vol.2717), pp.112-121. Springer-Verlag, June 2003.
  • Peter Rogelj, Stanislav Kovačič."Rigid multi-modality registration of medical images using point similarity measures". In: O. Drbohlav (ed.), Proceedings of the 8th Computer Vision Winter Workshop CVWW'03, pp.159-163. February 2003.
  • Peter Rogelj, Stanislav Kovačič."Validation of a Non-Rigid Registration Algorithm for Multi-Modal Data". In: Milan Sonka, J. Michael Fitzpatrick (eds.), Medical Imaging 2002, Image Processing : 24-28 February 2002, San Diego, USA, (Proceedings of SPIE, vol.4684), pp.299-307. SPIE, Bellingham (USA), February 2002.
  • R. Bajcsy, S. Kovačič. "Multiresolution Elastic Matching". Computer Vision, Graphics and Image Processing, 46: 1-21, April 1989.