Adaptive regularization for non-rigid image registration

A new adaptive regularization algorithm for non-rigid image registration has been presented. We firstly formulated the mathematic model of nonlinear registration. Then the local image information is incorporated into the mathematical framework and the amount of regularization changes adaptively according to image structure. As a result, the deformation can be smoothed selectively and the local detail of deformation was preserved. In order to reduce time complexity, additive operator splitting method is adopted to solve the partial difference equation. Experiments using synthetic images and MRI images show that the new method achieves better performance than the non-adaptive regularization methods. © 2012 American Scientific Publishers, All rights reserved.