A scale space method for volume preserving image registration

Abstract

Image registration is an inherently ill-posed problem. Therefore one typically aims to provide as much information about the under-lying application as possible. In particular for tumor monitoring, volume preservation of the wanted deformation is a central point. Based on [9], we propose a new scale space approach to volume preserving image registration. The main advantage of the new approach is that the constraints appear linearly and therefore the system matrices resembles Stokes matrices, which appears in computational fluid dynamics. We present the scale space framework, a composition based numerical approach and its implementation. Finally, we demonstrate the outstanding features of this idea by a real life example. © Springer-Verlag Berlin Heidelberg 2005.