[ITK-users] general info ICP implementation

[D'Isidoro Fabio]

I am trying to figure out how ICP is implemented in ITK. The general approach consists in iterating two steps: first compute the correspondences and then find the best transform with the current correspondences by using a close formula.

However in the ITK implementation I can't see this approach, plus LevenbergMarquardt alrgorithm is used instead of close formula to determine the transform. I know a special version of LevenbergMarquardt-ICP algorithm has been proposed in the following paper, however I can't see how the ITK implementation follows this proposed method (the ITK implementation seems to me very generic like any other optimization problem). Where should I look for?

Paper: "Robust registration of 2D and 3D point sets" Fitzgibbon 2003.

Thank you very much.
Fabio

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20150810/bae5f91a/attachment.html>