[Insight-users] penalizing the first derivative of displacements

[Anja Ende]

Hello everyone,

This is more of an image analysis question than an ITK question but I
am hoping someone here with more image analysis can help me.

I am coding up a simple registration application using ITK and I am
using the BSpline transformation model. What I initially want to do is
use the L2 norm of the difference in the displacement between
neighbouring control points a penalty term. However, I found out that
this of course seems to favour solutions where the spline points move
together (since then it vanishes). And I noticed that in the results
as the transformations I get seem to be translations of all the
control points together, which result in a simple translation of the
image.

I am aware of other reg. like bending energy of the spline and other
but was wondering if using penalty based on displacement of the
control points was in itself such a bad idea! I would be very glad if
someone could enlighten me on this!

Sincerely,

Anja