[Insight-users] point based registration using least square

[David Doria]

On Tue, May 4, 2010 at 8:56 PM, Suyang Mei <suyang_mei at yahoo.com> wrote:

> Hi, all -
>
>   I have a simple question here -
>
> I'd like to perform point based image registration by using least square
> approach (basically finding the rotation matrix by using singular value
> decomposition SVD). Inputs are two sets of corresponding points, and I just
> need to get the transformation matrix back.
>
> More like the functions in Examples\Patened\ICP*.cxx, but use closed form
> instead of iterative approach. What is the ITK filter to perform this?
>
>   Thanks a lot in advance.
>
> Suyang
>

I've never done this in ITK, but in VTK the ICP filter is the iterative
wrapper around vtkLandmarkTransform. So the ICP filter determines the
correspondences, then the closed form solution of that known correspondence
registration problem is done with the LandmarkTransform filter.

Maybe this will point you in the right direction? Or someone can point you
to something similar in ITK?

http://vtk.org/Wiki/VTK/Examples/Filters/LandmarkTransform
<http://vtk.org/Wiki/VTK/Examples/Filters/LandmarkTransform>
http://vtk.org/Wiki/VTK/Examples/Filters/IterativeClosestPointsTransform
<http://vtk.org/Wiki/VTK/Examples/Filters/IterativeClosestPointsTransform>
http://vtk.org/Wiki/VTK/Examples/Filters/ICPInputData
<http://vtk.org/Wiki/VTK/Examples/Filters/ICPInputData>
Thanks,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100505/5c2191f0/attachment.htm>