[Insight-users] Re: How to perform FFT on a huge dataset?

Luis Ibanez luis.ibanez at kitware.com
Thu Aug 3 17:29:48 EDT 2006


Hi GL

Analyze is probably better than DICOM for this particular
application.

Then you apply a process in two stages:

1) The easy / fast part
2) The hard / slow part



The Easy part involves to first export your volume into a
series of 2D Analyze files (one per slice).

You can do this with the ImageSeriesWriter, as explained in
the "Reading and Writing Images" chapter of the ITK Software
Guide:

         http://www.itk.org/ItkSoftwareGuide.pdf

Then for each one of those files you do the 2D FFT, and save
the results in a corresponding set of new 2D Analyze files.


The hard/slow part involves to read the first pixel from
each one of these N 2D files, put them in a linear array
and run a 1D FFT on it, collect this as the first row of
an image, then read the second pixel of all the 2D files,
put it into a linear array, compute its FFT and put the
result as the second row of an output 2D image.... you
will proceed as this until you fill that 2D image.

Note that the output set of 2D image will be a permutation
of the axis of the original image. Depending on what you
are going to do with that FFT, you may want to reorder
the pixels or not.


     Regards,


        Luis



-------------------
Guilian Xu wrote:
> Hi Luis,
> 
> We use Analyze or DICOM file, we can always translate to these two file
> format.
> 
> Thanks for your help,
> 
> GL
> 
> -----Original Message-----
> From: Luis Ibanez [mailto:luis.ibanez at kitware.com] 
> Sent: Wednesday, August 02, 2006 8:00 PM
> To: XU,GUILIAN
> Cc: Insight Users
> Subject: Re: How to perform FFT on a huge dataset?
> 
> 
> Hi GL
> 
> 
> In what fileformat is your current image stored ?
> 
> 
>    Please let me know,
> 
> 
>       Thanks
> 
> 
>         Luis
> 
> 
> --------------------
> XU,GUILIAN wrote:
> 
>>Dear Luis,
>>
>>How to only read in 1 slice or 1 line at a time, and send the FFT result 
>>back the 3D image correctly? Could you give me some directions/clues of 
>>how ITK can handle these easily.
>>
>>Thank you very much for your helps
>>
>>GL
>>
>>
>>
> 
> 
> 
> 
> 
> 




More information about the Insight-users mailing list