Elastic Image Registration using Parametric Deformation Models

                           Jan Kybic

            Robotvis, INRIA, Sophia-Antipolis, France
                         kybic@ieee.org

This talk presents an overview of the work described in my PhD thesis.
Its main topic is elastic image registration targeted at biomedical
applications. We start with an overview of the registration problem.
We show that the manual approach requires a landmark interpolation
that we generalize using a variational approach. This leads to some
interesting applications of variational reconstruction.

We then describe our general elastic image registration algorithm.  It
uses a grid of uniform B-splines to describe the deformation.  It also
uses B-splines for image interpolation. Multiresolution in both image
and deformation model spaces yields robustness and speed.  A
semi-automatic version of the registration algorithm is capable of
accepting expert hints in the form of soft landmark constraints.

We conclude with several examples of applications: EPI distortion
compensation, MRI atlas registration, CT alignment, Xenon SPECT atlas
creation, and echocardiography sequence analysis.