Re: [Insight-users] problem with registering a sequence of images

[Tomáš Kazmar]

Hi Chunfang,

  I think itk registration works with images according to their spacing. The problem
you face is maybe that of spacing and itk::WarpImageFilter (or how do you
transform the moving image?), maybe something like the following is what you are
looking for:

warper->SetOutputSpacing(movingImage->GetSpacing());
warper->SetOutputOrigin(movingImage->GetOrigin());

Look at the comments in itkWarpImageFilter.h for more information.

Regards,
Tomas


# Hi, Tomas
# 
# thanks for your kind help.
# I am so sorry to have not addressed my problem clearly.
# 
# my task is to register a sequence of plant slice images to construct a 3D
# modal of this plant.
# The plant slice images are taken separately and the deformation between them
# are created by the outside conditions.
# So in my work, I use the CenteredRigid transform modal to represent this
# deform.
# I find out that the spacing of the image I got after first registration is
# different from the original fixed image, so when this image is taken as the
# reference image in the second registration, the results turns wrong, which
# is a very small object  in the center.
# I want to ask you that why the registration in ITK change the spacing of the
# image, or if I got some error in understanding this problem.
# 
# The suggestion in your letter is very useful, and I should choose a pivot
# image to align all the other to it
# I have read the groupwise registration method you recommanded to me, but I
# got confused that wether this groupwise method suits my situation.
# 
# 
# Best wishes,
# chunfang
# 
# 2008/2/15, Tomash.Kazmar at seznam.cz <Tomash.Kazmar at seznam.cz>:
# >
# > Hi Chunfang,
# >
# > you do not provide sufficient information to tell why your algorithm
# > fails.
# > Did you check the tranfsormation you get after first registration and
# > the resulting image. How wrong the result of the second registration is?
# > Which kind of transformation between images do you wnat to recover?
# > Please tell us the details and add the relevant part of the source.
# >
# > Another thing is that the algorithm as you describe it, will accumulate
# > the errors. What is your task? Do you have a finite sequence? Is not it
# > better to choose a pivot image and align all other images to it, or to use
# > groupwise registration:
# >
# >
# http://www.insight-journal.org/InsightJournalManager/view_publication.php?pubid=173
# >
# >
# > Regards,
# > Tomas
# >
# > # there is a sequence of images, what I want to do is to take the first
# > one as
# > # the reference, warp the second one using registration, then take the
# > warped
# > # second image as the new referece image to register the third one, and it
# > # goes on like this
# > #
# > # the processe goes well when registering the original second image using
# > the
# > # first image as refernce,
# > # but when the registed second image was used as the reference image to
# > # register the third image in the sequence, the result turns wrong.
# > # the result image is very small, the
# > #
# > # I wonder whether the image created by the ITK registration has some
# > # difference
# > # can someone help me to find the problem, I get confused about this
# > problem
# >
# 
# 
#