In this registration toolkit we also propose a non-linear solution for mono-modal image registration. This method provides a very up-to-date method [#!Vercauteren:ISBI:2006!#] based on variants of the Thirion's demons algorithm [#!Thirion:MIA:1998!#]. Although this procedure takes longer to converge than the affine automatic mode, it gives consistent results for inter-patient data. The output is a deformation vector field. We want to thank Tom Vercauteren and Mauna Kea Technologies for their participation on this registration algorithm.
You can save the resulting deformation field in ITK metafile format (.mha) thanks to this button. The deformation field maps the moving image onto the fixed image. A deformation field is represented as an image whose pixel type is some vector type with N elements, where N is the dimension of the fixed image.
With this other button you can load a previously saved deformation field. It currently takes only ITK metafile vector images.
IMPORTANT : This is a known issue that linear transformations (i.e. rigid or affine) are not linked with non-linear ones (diffeomorphic demons) : Considering a fixed image , a moving image , if one performs a rigid transformation, resulting in an image (and an output affine matrix ), and then performs a non-linear diffeomorphic registration, resulting in an image and a deformation field , then the output deformation field corresponds to the deformation of the ``mid-way image'' onto . Another known issue is that you currently cannot save and load deformation field for 2D image non-linear registration.