[Insight-users] Re: Optimzers and metrics for point set to point

[Harry Hoellig]

Hi,

thank you very much for your answer.

Indeed I'm using the ICP (I should have explained it more deeply), but 
instead of the translation transform used in the example to perform the 
registration I want to use several transforms with more degrees of 
freedom like rigid, similarity and affine transforms.

So when I simply change the transform type to these tranforms, the 
LevenbergMarquardtOptimizer seems to destroy the transform matrices. 
That's why I'm looking for optimizers that are capable of optimizing 
these transforms (I only found the VersorRigid3DTransformOptimizer with 
the problem stated before)...So, are there any optimizers which I can 
use with these transforms?

Thanks,
Harry


Markus Weigert schrieb:
> Hi Harry,
>  
> why don't you think about using iterative
> closest point algorithm (ICP) for the task of point set registration?
> I think this should be the simplest solution...
>  
> See also
> http://voxel.jouy.inra.fr/darcs/itk/Insight/Examples/Patented/IterativeClosestPoint3.cxx
>  
>  
> Regards,
> Markus