Zernike Moment and Mutual Information Based Methods for Multimodal Image Registration

Abstract

Image registration enables joint operations between images obtained from diverse sources. However, there have been limited advances in the registration of multichannel images. The accuracy of registration is a significant concern for medical applications, among others. Two methods, PCA–ZM and CED–ZM, have been proposed for registration based on Zernike moment and enhanced mutual information. Edge detection by Zernike moment and identification of common features in multichannel images are used as a foundation to improve accuracy over single-channel registrations. Single-channel registration accuracy for MRI and SPECT brain images is found to surpass the methods compared against. PCA–ZM demonstrates good accuracy for MR-MR registration, while CED–ZM has good accuracy for MR-SPECT registration. These measures improve upon accurate registration for images, especially where many modalities are available, such as in medical diagnosis.