[Insight-users] Landmark Fitting
Luis Ibanez
luis . ibanez at kitware . com
Fri, 16 Aug 2002 11:17:36 -0400
Hi Xenios,
I missunderstod your question,
Your interest is actually on PointSet to PointSet registration.
We don't actually have this family of methods in ITK.
It will be quite useful to have them in ITK.
If you are willing to contribute these methods to ITK
there is a variety of ways of doing so.
One interesting option is the current RFQ of the NLM by which
you can get funding for porting existing algorithms into ITK:
http://www.nlm.nih.gov/oam/oam.html
Please look under:
RFQ NLM 02-148/KDR for "Algorithm, Adapters and
Data Distribution for the Visible Human Project
Insight Toolkit"
The deadline is August 30th but the process of submission is
quite simple.
Of course you may also contribute code without being funded
by the NLM.
Thanks
Luis
===============================================================
Xenios Papademetris wrote:
> Luis,
>
> Thanks for the response. I think you misunderstood my question (probably
> my fault), so let me try again:
>
> Given a point set A and a corresponding point set B (let's arbitrarily
> assume each set has N points) are there methods in ITK for
>
> 1. Fitting the best affine transformation between A and B (c.f.
> vtkLandmarkTransform in vtk)
>
> 2. Fitting the best spline transform with M (M not equal to N) control
> points using the point sets A and B.
>
> The methods that are in ITK seem to primarily do interpolation, I guess
> our interest is more in terms of approximating the correspondences with
> a more compact (or generic) transformation.
>
> We are in the process of designing an implementation of a set of robust
> point-based matching (RPM) registration algorithms in ITK that came out
> of work done earlier at Yale by Chui, Rangaranjan and collaborators over
> the last 7-8 years. These can be seen as a generalization of the
> Iterative Closest Point algorithm (ICP) -- indeed with a specific choice
> of parameters RPM reduces to ICP.
>
> Thanks again,
>
> Xenios
>
>>