ITK  4.2.0
Insight Segmentation and Registration Toolkit
Classes | Namespaces
itkFEMRegistrationFilter.h File Reference

Go to the source code of this file.

Classes

class  itk::fem::FEMRegistrationFilter< TMovingImage, TFixedImage, TFemObjectType >
 FEM Image registration filter. The image registration problem is modelled here with the finite element method. Image registration is, in general, an ill-posed problem. Thus, we use an optimization scheme where the optimization criterion is given by a regularized variational energy. The variational energy arises from modeling the image as a physical body on which external forces act. The body is allowed to deform so as to minimize the applied force. The resistance of the physical body to deformation, determined by the physics associated with the body, serves to regularize the solution. The forces applied to the body are, generally, highly non-linear and so the body is allowed to deform slowly and incrementally. The direction it deforms follows the gradient of the potential energy (the force) we define. The potential energies we may choose from are given by the itk image-to-image metrics. The choices and the associated direction of descent are : Mean Squares (minimize), Normalized Cross-Correlation (maximize), and Mutual Information (maximize). Note that we have to set the direction (SetDescentDirection) when we choose a metric. More...

Namespaces

namespace  itk
namespace  itk::fem