Covariance matrix based elastic multi-channel image registration

G. K. Rohde, S. Pajevic, C. Pierpaoli, P. J. Basser National Institutes of Health, Bethesda, Maryland, United States SYNOPSIS: With the advent of new MR imaging modalities such as DT-MRI, MRA, and CSI, demand for image registration procedures capable of dealing with multi-channel image data has increased. A novel method based on multivariate linear correlation is proposed to align two multi-channel images each containing an arbitrary number of channels. Results obtained using diffusion tensor MRI data of the human brain show that image registration based on different channels generates different alignment results. In addition, we show that multi-channel image registration is more accurate to registration based on a single channel.