[Insight-users] VNLFFT vs. FFTW

[Kevin H. Hobbs]

I'm doing

1.
read -> FFT -> FFTShift -> modulus -> log -> file1

Using the file to make a mask and then doing:

2.
read_mask -> FFTShift
                     \
                      --> mask -> iFFT -> file2
                     /
          read -> FFT

When I use FFTW I have to do UpdateLargestPossibleRegion() at the end of
1. or the mask in 2. fails complaining about the regions being at least
partially outside.. well duh FFTW uses symmetry to only make half the
output. But then file2 gets doubled along the x-axis with vertical lines
of junk (I'm surprised it worked at all).

I switched to using the VNL classes and all is well, I suppose I could
crop the masked image if I know that the iFFT is the FFTW class... but
that seems like a bit of a kludge.

It should probably be a bug that a program writen with
itk::FFTRealToComplexConjugateImageFilter and
itk::FFTComplexConjugateToRealImageFilter will fail based on weather itk
was made with FFTW.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 253 bytes
Desc: OpenPGP digital signature
URL: <http://www.itk.org/pipermail/insight-users/attachments/20110802/c7dfc78b/attachment.pgp>