Embedded Entropy-Based Registration of CT and MR Images

Abstract

In this paper, a novel approach has been developed for registration of noisy MR images with CT images. Noisy MR image is considered as floating image and CT image as the reference image. Here, the structural information of the images has been presented for both the images. The structural representation of the images has been computed by embedding Arimoto entropy and Tsallis entropy. MR image is the noisy one, and hence, Gaussian kernel filter is applied to estimate the pixel for structural representation. These structural images are considered for image registration. The joint histogram of structural images has been computed and then is used to compute mutual information (MI) which is considered as the similarity metric for registration framework. Our proposed method has been analyzed for different noise (SNR) conditions and also with other methods from literature.