A novel dynamic multilevel technique for image registration

We present a novel dynamic multilevel technique for solving image registration problems. The development is carried out to construct a spatial transformation based on cubic B-spline basis functions and determine the control points dynamically. Unlike FEM-based image registration methods, we do not have the difficulty of solving a complicated matrix system. In addition, the presented method is enhanced by a multilevel technique, which makes it more efficient and flexible. The numerical results and several comparison studies on real bio-medical images show our technique is stable, accurate and fast, especially for large deformation registration problems.