Hi Jan,
The T1 image is likely a 3D image the DWI image is a 4D image. You’d need to reduce the DWI image to a 3D image for instance by averaging across volumes (mrmath b0_in.mif mean -axis 3 b0_out.mif).
However, I doubt that registration will work with mrregister. Have a look in the forum for alternatives. For instance: Registration of structural and diffusion weighted data - #7 by ElijahMak