[Insight-users] Which Kernel transform to use with Point Set Registration

[Luis Ibanez]

Hi Aya,


        The KernelTransforms are Transforms in themselves.

They map points for you by using the TransformPoint() method.


The space transformation produced by a KernelTransform has
more degrees of freedom than the transformation produced by
an Affine transform.  Therefore, it is not possible to summarize
the KernelTransform with the typical 4x4 or 4x3 matrices that
tend to be used for representing Transformations.


If you want to restrict the transformation to an Affine or Rigid
transform, then what you probably want to use is the class


       Insight/Code/Common/
                   itkLandmarkBasedTransformInitializer.h
                   itkLandmarkBasedTransformInitializer.txx


In this class, you are also required to provide two collections
of points, each one with the same number of points, and for
which you have established correspondences.


     Regards,


           Luis


-----------------------------------------------------------------------------------------
On Thu, Feb 18, 2010 at 10:28 AM, Aya El Gebeely
<aya.elgebeely at symbyo.com> wrote:
> Hi Luis, Thanks a lot for your reply..
>
> I can have the same number of point set for the both images, so I selected
> itkThinPlateSplineKernelTransform.h to use as my transform method..
>
> It gives me the mappedPoints, but what I need is the transformation values
> [i.e Transform along X and Y] as well as the Transformation Matrix..
>
> To be able to apply the transformation on the image we want to register..
>
> I'll be glade to show me how I could get these values ..
>
> Thanks in advance,
> Aya R.
>
>
> On 18 February 2010 01:54, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>>
>> Hi Aya,
>>
>>
>> Do you need a Deformable Transform ?
>>
>>
>> Typically you should start a point set to point set registration
>> by just using a Rigid Transform.  For example by using
>> ICP (iterative closest points), as described in the examples:
>>
>> Insight/Examples/Registration/
>>    IterativeClosestPoint1.cxx
>>    IterativeClosestPoint2.cxx
>>    IterativeClosestPoint3.cxx
>>
>>
>> ....
>>
>> but..
>>
>> Should you really need to use KernelTransforms, please note
>> that they will require you to provide point correspondences.
>>
>> That is, you should know what point of the FixedCollection of
>> points, correspond to what point of the MovingCollection of
>> points.
>>
>> If you have such correspondences, then you could use any
>> of the following KernelTransforms:
>>
>>       itkThinPlateSplineKernelTransform.h
>>       itkThinPlateR2LogRSplineKernelTransform.h
>>       itkElasticBodySplineKernelTransform.h
>>       itkElasticBodyReciprocalSplineKernelTransform.h
>>       itkVolumeSplineKernelTransform.h
>>
>>
>>
>>     Regards,
>>
>>
>>             Luis
>>
>>
>>
>> ------------------------------------------------------------
>> On Wed, Feb 17, 2010 at 9:49 AM, Aya El Gebeely
>> <aya.elgebeely at symbyo.com> wrote:
>> > Dear ITk Users,
>> >
>> >  Currently I'm using Point Set to Point Set registration method, in
>> > order to
>> > register two medical images, given set of points for each image [fixed &
>> > moving]
>> >
>> > I need to use KernelTransforms, so which one to use, it was mentioned in
>> > the
>> > ITk S/W guide to use any of the itk::KernelTransform family of classes..
>> >
>> > Do you have any recommendations ?
>> >
>> > Thanks in advance,
>> >
>> > Aya R. ElGebeely
>> >
>> > _____________________________________
>> > Powered by www.kitware.com
>> >
>> > Visit other Kitware open-source projects at
>> > http://www.kitware.com/opensource/opensource.html
>> >
>> > Kitware offers ITK Training Courses, for more information visit:
>> > http://www.kitware.com/products/protraining.html
>> >
>> > Please keep messages on-topic and check the ITK FAQ at:
>> > http://www.itk.org/Wiki/ITK_FAQ
>> >
>> > Follow this link to subscribe/unsubscribe:
>> > http://www.itk.org/mailman/listinfo/insight-users
>> >
>> >
>
>