High b-valued diffusion-weighted images (DWI), which were designed to solve fiber-crossing problems, are susceptible to many artifacts and distortions. Since DWIs with different diffusion gradients produce dissimilar intensity contrasts, and since the distortion is nonlinear when multiple artifactual sources are intermixed, the mutual information-based affine registration may not be adequate for precise correction of distortions in DWIs, especially for images acquired with high b-values. To overcome these problems, we proposed an iterative image registration technique through which simulated DWIs are generated, driven from a diffusion tensor estimate, as targets for measured DWIs in the registration. Since simulated DWIs have similar intensity profiles to those of measured DWIs and the same geometric profiles as b(0)-images, an iterative procedure enables intensity-based nonlinear registration. As a pre-processing step, we also proposed a motion detection and sub-volume utilization for interleaved volumes. Performance evaluation with high b-valued DWIs for high angular resolution diffusion imaging and diffusion kurtosis imaging showed that the proposed method had a superior advantage over the conventional registration technique