Cumulative Gausian Optimizer for Registration

Michael7 · January 18, 2021, 10:04am

Hi,

I’m trying to configure and set an itk::CumulativeGaussianOptimizer instance to run a registration ( itk::PointSetToPointSetRegistrationMethod<PointSetType, PointSetType >) procedure.
Nevertheless, I seem to find no example in the documentation, on the web, in the itk Manual nor in the itk forums.
I was wondering if somebody could point me towards an example if possible like the ones we can find concerning the itk::LevenbergMarquardtOptimizer.

Best Regards,

Michael O.

dzenanz · January 18, 2021, 10:52pm

I think this is the example you are looking for:

github.com

InsightSoftwareConsortium/ITK/blob/v5.2rc01/Examples/RegistrationITKv4/IterativeClosestPoint3.cxx#L152-L159


using OptimizerType = itk::LevenbergMarquardtOptimizer;
OptimizerType::Pointer optimizer = OptimizerType::New();
optimizer->SetUseCostFunctionGradient(false);
// Registration Method
using RegistrationType =
  itk::PointSetToPointSetRegistrationMethod<PointSetType, PointSetType>;

dzenanz · January 18, 2021, 10:56pm

I don’t think there is an example for CumulativeGaussianOptimizer, in case you were asking for an example similar to the one I provided in previous post.

Michael7 · January 20, 2021, 10:25am

Hi, thanks for your answer, but I’m indeed looking for a specific CumulativeGaussianOptimizer example used at a registration procedure. Since the optimizer (LM) presented in the IterativeClosestPoint3.cxx does not have the same configuration and settings I’m not able the run it when changing the optimizer type (obtaining a segfault).

dzenanz · January 20, 2021, 2:12pm

If you decide to go through with using CumulativeGaussianOptimizer and get it working, it would be good if you contributed it as a new example for the benefit of others.

Michael7 · January 21, 2021, 8:41am

I would like to ! I think I’ll post asap an unitary test containing a simple mesh loader and registration procedure with vtk/itk hoping that we could together -with every little advice of any interested person- collaborate to make it work.