Noisy brain MR and CT image registration using MRF model

In this paper, a novel approach has been proposed to register noisy brain multimodal images. Specifically, we have considered CT and MR images. The MR image is the noisy one that has been considered to be the floating image while CT image is the fixed one. The joint histogram of the CT and noisy MR image has been modeled as Markov Random Field (MRF). Every element of the joint histogram has been estimated using the Maximum a Posterior (MAP) estimation criterion. The MAP estimates have been obtained jointly by Simulated Annealing (SA) and Iterative Conditioned Mode (ICM) algorithms. This estimated joint histogram has been used to compute the mutual information for different values of the registration parameter. The registration parameter has been estimated with 99 percent confidence interval. The proposed scheme has been successfully tested with different noise conditions and has been compared with five existing methods.