Spatial mutual information for non rigid registration.

With current state of the art approaches, both the similarity measure and transformation model are generally the same over the whole image. However, it is clear that both the required regularisation of the transformation as well as the available image based registration clues form which to assess the image similarity are spatially varying, for instance anatomical objects or tissues having different rigidity, homogeneous versus contrasting image regions, or image ambiguity due to noise, intensity inhomogenity, ... . Within this research, we want to extend the current available paradigm for MI-based image registration with the incorporation of spatial information, both in the similarity measure as in the transformation model. We will call this new approach 'spatial mutual information' (sMI).

Keywords:Mesh, Mutual information, Image registration, Medical image computing, Free-form deformation, Fusion, Warping

Disciplines:Medical imaging and therapy, Other paramedical sciences, Multimedia processing, Biological system engineering, Signal processing