[Insight-users] BSpline - deformable transformation with default multiresolution framework

[Luis Ibanez]

Hi Reinhard,

I may not have been clear enough, I apologize for that.


The BSpline deformable transform itself *IS* compatible
with the Multi-Resolution framework in ITK.

What I meant in my previous email is that the approach
of creating *multiple* BSPline transforms, by changing
their grids resolution is not something you can plug
directly into the Multi-Resolution framework.

In principle you could change the Transform in the middle
of the registration process, e.g. by taking advantage of
an Iteration event, as you suggested. However, you will have
to be careful to make sure that the array of parameters that
is passed to the optimizer and to the Metric is resized
correctly in order to reflect the change in number of
parameters from the Transform.


    Regards,


       Luis



=========================
Reinhard Hameeteman wrote:
> Hi Luis and Markus,
> 
> Now this last post worries me a bit. I don't really see a reason why
> the B-spline transform shouldn't work with the multiresolution
> framework. But I think you both are referring to up-sampling the node
> density of the transform when going to a higher resolution level,
> which is not done in the multiresolution framework right? But since
> the first thing done in the registration is invoke the iteration event
> it should very wel be possible to up-sample the grid at that point,
> wouldn't it?
> 
> Regards,
> Reinhard
> 
> 2006/4/30, Luis Ibanez <luis.ibanez at kitware.com>:
> 
>>
>> Hi Markus,
>>
>> You are correct in your observation.
>>
>> The approach of using a multi-resolution BSpline grid is a different
>> mechanism than the one used in the standard multi-resolution framework.
>>
>> In this case, as you pointed out, we are refining the transform itself.
>>
>> At this point there is no direct way of integrating the BSpline multi
>> resolution with the pyramidal framework for multi-resolution
>> registration. It is, however, a very interesting undertaking and it will
>> be worth to invest some time on it if you find that in alignment with
>> your current work.
>>
>> Please let us know if you need any help with the details of setting
>> up such a framework.
>>
>>
>>
>>     Thanks
>>
>>
>>        Luis
>>
>>
>> ========================
>> Markus Weigert wrote:
>> > Hi all,
>> >
>> > I had a look at the multiresolution example of BSpline - registration
>> > (example 6)
>> > and noticed, that using BSpline in this way doesn't conform to the
>> > multiresolution
>> > registration framework (especially, it uses two transform objects and
>> > itkImageRegistrationMethod
>> > instead of using itkMultiResolutionImageRegistrationMethod).
>> > My question is now:
>> > Is there a way to integrate this transform in the multiresolution
>> > framework to use it like any other transform,
>> > or do I have to change my code to deal with different registration
>> > schemes???
>> > My goal is a class, which uses the multiresolution framework and
>> > provides several metrics, optimizers,
>> > transforms and interpolators in a unified way.
>> >
>> > Regards,
>> > Markus
>> >
>> >
>> > 
>> ------------------------------------------------------------------------
>> >
>> > _______________________________________________
>> > Insight-users mailing list
>> > Insight-users at itk.org
>> > http://www.itk.org/mailman/listinfo/insight-users
>>
>>
>> _______________________________________________
>> Insight-users mailing list
>> Insight-users at itk.org
>> http://www.itk.org/mailman/listinfo/insight-users
>>
> 
>