[Insight-users] Follow up: About the up-sampling of deformation field from low resolution to high in Examples/DeformableRegistration6

[Mengda Wu]

Hi all,

    I just would like to follow up with my last email. I think the
up-sampling code (line 340 to 375) is wrong.

   Since the ResampleImageFilter is operated on the coefficients (not the
deformation field mentioned in my last email), what its output should
be already coefficients (not the deformation field) in higher resolution.
Using BSplineDecompositionImageFilter to get the coefficients again
does not make sense.

   For reference, a method for up-sampling of the deformation field is
mentioned in Mattes's 2003 paper, in particular equations (17) and (18).

Thanks,
Mengda

---------- Forwarded message ----------
From: Mengda Wu <wumengda at gmail.com>
Date: 2010/4/1
Subject: About the up-sampling of deformation field from low resolution to
high in Examples/DeformableRegistration6
To: Insight-users at itk.org


Hi all,

   I have a question about conversion of deformation field from low
resolution to high in Examples/DeformableRegistration6. I do not understand
why we need both ResampleImageFilter and BSplineDecompositionImageFilter. I
think we only need to up sample the deformation field (per dimension)
using ResampleImageFilter with BSplineResampleImageFunction.

  Then, why is the output of ResampleImageFilter fed into
BSplineDecompositionImageFilter and then put into the parameters? What does
BSplineDecompositionImageFilter do?
I cannot get this from the document of BSplineDecompositionImageFilter.

Thanks,
Mengda
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100401/1943bd28/attachment-0001.htm>