[Insight-developers] Please help mewithBSplineDeformableTransformation

[Stefan Klein]

Hi,

The B-spline control points outside the image are added with a reason. They 
influence the deformation of pixels at the edge of the image. If you force 
them to be zero, you somehow constrain the deformation field at the edge of 
the image. This might not at all be what you want, unless you are sure that 
the deformation at the edge of the image approaches zero. In the case of a 
translation of a part of the image close to the edge this assumption would 
for example already be violated.

Regards,

Marius and Stefan.


At 11:26 21/03/06, you wrote:
>There a number of ways to this.  The one off the top of my head requires
>writing a new BSplineDeformableTransform subclass.
>
>If the control points outside the grid are to be fixed in location,
>then those control points could be stored in the transform's 
>"FixedParameters".
>The "Parameters" for optimization would be just the control points over 
>the domain
>of the image.
>
>Transforming a point with the new transform is now much more complicated 
>because some
>of the spline coeffiecients are in the fixed parameters and some are in the
>parameters to be optimized.  One could keep an extra set of "full control 
>points"
>and copy the parameters that are optimized into the full set of control 
>points as
>needed, then use the current code for evaluating the deformation.  This is 
>relatively
>simple to do but would be very slow in practice. Or, the transform code 
>could be written
>such that information was pulled from the fixed parameters or from the 
>parameters to be
>optimized as needed to evaluate the current position in the B-Spline. This 
>would probably
>be a tad bit slower than the original code for evaluating the transform, 
>due to some
>extra conditionals, but would be faster than the first option.  However, 
>it requires
>writing new code for evaluating a BSpline.
>
>Another option is write a whole new optimizer that will actually avoid 
>using the
>fixed control points in the any of the calculations of the metric 
>dervitive, etc.
>
>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: Tuesday, March 21, 2006 10:53 AM
>To: insight-developers at itk.org
>Subject: Re: [Insight-developers] Please help
>mewithBSplineDeformableTransformation
>
>
>Please help, isn't there a way? It has to be useful for every kind of
>flexible registration using BSplineDeformableTransform.
>
>
>
> > To have a switch for optimization would be very useful. Does anybody
> > have an idea? How complicated would a wrapper for
> > BSplineDeformableTransform be?
> >
> > Henning
> >
> > 2006/3/20, Miller, James V (GE, Research) <millerjv at crd.ge.com>:
> > > I think for the LBFGSBOptimizer, you use the SetBoundSelection() 
> method along with the SetLowerBound(), SetUpperBound() methods.
> > >
> > > Jim
> > >
> > > -----Original Message-----
> > > From: Henning Meyer [mailto:tutmann at gmail.com]
> > > Sent: Monday, March 20, 2006 9:25 AM
> > > To: Miller, James V (GE, Research)
> > > Subject: Re: [Insight-developers] Please help me
> > > withBSplineDeformableTransformation
> > >
> > >
> > > How would one mark parameters for the optimizer, to tell it not 
> optimize those?
> > >
> > > Henning
> > >
> >
>_______________________________________________
>Insight-developers mailing list
>Insight-developers at itk.org
>http://www.itk.org/mailman/listinfo/insight-developers
>_______________________________________________
>Insight-developers mailing list
>Insight-developers at itk.org
>http://www.itk.org/mailman/listinfo/insight-developers

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.itk.org/mailman/private/insight-developers/attachments/20060321/6abb5cdd/attachment.html