[Insight-users] itkCurvatureRegistrationFilter

[Markus Weigert]

Hi insight users,

I currently tried to use the ITKCurvatureRegistrationFilter
with different registration functions (itkMIRegistrationFunction, 
itkSymmetricForcesDemonsRegistrationFunction, MeanSquaresRegistrationFunction)
to register two head/neck CT images.
I used a constraint weight of 1000 and a time step of 100 for 10 iterations (just to see
in which direction the deformationfield evolves).
The 2 CT images also show a slight rigid mis-registration, which curvatureRegistration
is said to take into account.

As compared to using itkSymmetricForcesDemonsRegistrationFunction
or itkMeanSquaresRegistrationFunction, using itkMIRegistrationFunction computes a 
displacement field which is opposed to the deformation field, using the other two metrics
(which apparently calculate a displacement field, that evolves in the right direction).

Following the sourcecode documentation of ITKCurvatureRegistrationFilter (ITK v2.4.0, I also
noticed that the source comments do not appear in the cvs doxygen documentation!!!), 
it should work with any RegistrationFunction.
I also tried to set DoInverse(true) on the itkMIRegistrationFunction, but still get the same results.
I guess it has something to do with the negative sign of the MI Metric but could still not figure it out.
Has somebody had the same effect or can comment on that???


Regards,
Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20061006/6cb4c2b2/attachment.htm