In this paper an automatic multiscale feature-based rigid-body coregistration technique for diffusion tensor imaging (DTI) based on the local curvature K and torsion r of the white matter (WM) fiber pathways is presented. As a similarity measure, the mean squared difference (MSD) of corresponding fiber pathways in (K, tau)-space is chosen. After the MSD is minimized along the arc length of the curve, principal component analysis Is applied to calculate the transformation parameters. In addition, a scale-space representation of the space curves is incorporated, resulting in a multiscale robust coregistration technique. This fully automatic technique inherently allows one to apply region of interest (ROI) coregistration, and is adequate for performing both global and local transformations. Simulations were performed on synthetic DT data to evaluate the coregistration accuracy and precision. An in vivo coregistration example is presented and compared with a voxel-based coregistration approach, demonstrating the feasibility and advantages of the proposed technique to align DT data of the human brain.