Research of geometric registration method for multi-modality image

To increases the efficiency of multi-modality medical image registration, a novel multiscale registration framework is proposed based upon an edge preserving total variation L1 norm scale space representation in this paper. Based on selecting edges and contours of an image according to the geometric size rather than the intensity values of the image features, the scale space is constructed, which ensures more meaningful spatial information for MI based registration. Moreover, an optimal estimation of the total variation L1 parameter is designed in this framework by training and minimizing the transformation offset between the images for automated registration. The new method is validated through simulated mono- and multimodal medical datasets with ground truth and temporal clinical studies from a combined PET/CT scanner.