[Insight-users] Vnl fft algorithm

[Renaud Isabelle]

Hi,
   
  Sorry to bother you again with the computation of FFT but I am really trying to understand.
   
  Actually, I computed the FFT of a 2D matrix with Matlab of one side and ITK on the other side. Then, I have tried to compare the results but I didn't succeed. Yet, I am sure that my inputs of my FFT are the same but despite this, I really don't have the same results, even if I consider a scaling factor.
   
  - First of all, I am looking for the zero frequency, equalled to the sum of the elements of my input matrix. In matlab, the zero frequency, is the first element returned at the upper left corner of the matrix. But, with the itk Vnl FFT, this isn't corresponding. 
   
  - I have taken a closer look at the vnl fft algorithm. itkVnlFFTRealToComplexConjugateImageFilter calls LegalDim() function that seems to return an error message if one dimension is not of size of a power of 2. However, while computing the Vnl FFT of a 2D matrice of size 80*20, I got no error message returned. 
   
  Taking a look at the functions vnl_fft_1d and vnl_fft_2d, it seems that fft_1d requires a size of power of 2. 
   
  So my questions is:
  - does itkVnlFFTRealToComplexConjugateImageFilter need an input image of size of power of 2? If so, why don't I have any error while executing. 
   
  - BTW, I read that itkVnlComplexConjugateToRealImageFilter returns not the ifft of a complex image but the absolute value of this ifft. Is it still true? Or is there any modification supplied for that? 
   
  I really hope that someone could help me to understand vnl FFT. 
   
  Isabelle
   
   

		
---------------------------------
 Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger
 Téléchargez le ici !  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20051118/aaa8c221/attachment.html