Mathematics Meets Medicine: An Optimal Alignment
Image registration is a central problem in image processing. The objective is to automatically establish correspondences between two images. Particularly in medical imaging, registration is essential to motion correction, treatment verification, navigation, or image fusion, the integration of information from different devices.
Today's most general approaches are based on a variational setting that allows a combination of different building blocks like distance measures, regularizer, and constraints. Central problems in registration are the identification of application conform building blocks, its ill-posedness, the efficient numerical treatment, and the validation of registration.
The talk introduces to image registration and addresses the above problems exemplarily on a variety of real-life examples. Emphasis is given to numerical approaches that lead to high dimensional constrained optimization problems.
Jan Modersitzki, McMaster University, Canada