Structural Representation: Reducing Multi-Modal Image Registration to Mono-Modal Problem

Registration of multi-modal images has been a challenging taskdue to the complex intensity relationship between images. Thestandard multi-modal approach tends to use sophisticated similaritymeasures, such as mutual information, to assess the accuracyof the alignment. Employing such measures imply the increase inthe computational time and complexity, and makes it highly difficultfor the optimization process to converge. The presented registrationmethod works based on structural representations of imagescaptured from different modalities, in order to convert the multimodalproblem into a mono-modal one. Two different representationmethods are presented. One is based on a combination ofphase congruency and gradient information of the input images,and the other utilizes a modified version of entropy images in apatch-based manner. Sample results are illustrated based on experimentsperformed on brain images from different modalities