[Insight-developers] Please help me withBSplineDeformableTransformation

[Miller, James V (GE, Research)]

I think these control points are marked as variables not to optimize 
when the LBFGSB optimizer is used. So it may not be impacting the optimizer
as much as one would expect.

Lydia may provide more insight on this.


Jim

-----Original Message-----
From: insight-developers-bounces+millerjv=crd.ge.com at itk.org
[mailto:insight-developers-bounces+millerjv=crd.ge.com at itk.org]On Behalf
Of Henning Meyer
Sent: Monday, March 20, 2006 8:40 AM
To: insight-developers at itk.org
Subject: [Insight-developers] Please help me
withBSplineDeformableTransformation


Hello,

I posted this question a while ago, but nobody cared about it. So I
simply try it again:
The BSplineDeformableTransformation needs additional grid-points in
every dimension according to the order of the spline used. So when
trying to register a 3D image with a 5x5x5 grid of a 3rd-order BSpline
I end up with (5+3)^3 *3 = 1536 parameters instead of 5^3 * 3 = 375
parameters which is a huge difference for the optimizer.
But normaly these border-grid-points remain zero anyway. So... is
there an easy way to simply get rid of these points (at least to tell
the optimizer not to care about them)?

This could improve optimization performance quite a lot.

Or did I misunderstand something dramatically?

Please help me,

Thanks

Henning
_______________________________________________
Insight-developers mailing list
Insight-developers at itk.org
http://www.itk.org/mailman/listinfo/insight-developers