[Insight-users] Statistical Shape Model - Principal Component Analysis in ITK

[Paolo Taboga]

Hi everyone,

I am trying to register to point sets with ITK, but I want to use a custom-made transformation.

I created a statistical shape model of a 3d mesh, that means I have a set of 3d point coordinates (a "mean" mesh) plus a matrix of eigenvectors representing the principal modes of variation of said mesh. Said principal modes are the only allowed deformable transformations of the mean mesh and can be seen as a deformation field.
Every eigenvector can be weighted in order to achieve different deformations and my goal is to find the right weights in order to register the model with a generic cloud of points.

The only transformations I want to use are:
-a rigid transformation (already implemented in itk::Euler3DTransform)
-the principal modes of variation I calculated in the above-mentioned matrix

The "pseudo code" I want to implement is:

1. register the 2 point sets with a rigid transformation (find translation and rotation that best match the mean mesh with the point cloud)
2. register again the 2 point sets with the deformable transformation (find the weights of every eigenvector to improve the match)
3. repeat steps 1-2 until convergence

Is there a way to "create" a custom-made transformation and to pass it to "itk::PointSetToPointSetRegistrationMethod"?

Thanks in advance,
Paolo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20081015/2bb62791/attachment.htm>