Textural mutual information based on cluster trees for multimodal deformable registration

摘要: Mutual information (MI) has been widely used in image analysis tasks such as feature selection and registration. In particular, it is the most similarity measure for intensity based registration of multimodal images. However, a major drawback MI that does not take spatial neighbourhood into account. An effective way incorporating could be great benefit number challenging applications. We propose use cluster trees to efficiently incorporate textural from local voxel computation MI, while at same time limiting bins represent this higher-order information. This new metric optimised using Markov random field (MRF). apply our method dynamic lung CT volumes with simulated contrast. Experimental results show advantages technique compared standard mutual