[Insight-users] Question: Registration of vectorfields

[Markus Weigert]

Dear all,

I have a question concerning the transformation of vector fields.
Assuming, I would have two 3D vector images which I would like to register 
rigidly.

What would happen to a vector image after a transformation (say rotation + 
translation) is applied?
Would also the orientation of the vectors (image information) change due to 
the  transform (that is what I want) or only the position of the vectors?
Of course, this would rise the need to treat voxels like spatial objects which 
have an orientation.

Furthermore:
How would the deformable registration algorithms in ITK deal with two vector 
images as input? Do you know any metric which could be applied for such a 
task? I think, this would rise the need for some new kind of deformation 
model where the transform for ech voxel is not only a translation but some 
kind of rigid transformation and the metric would also need to take the 
orientation into account.
However, the question for me is if something like that would still be 
computable on a common personal computer.
These questions arrose from meg where image information is a vector field
and needs to be transformed for example to some MR images.


Best regards,
Markus