Re: Can't Mask FFT

Posted by Michael Schmid on Oct 31, 2008; 4:36pm
URL: http://imagej.273.s1.nabble.com/Can-t-Mask-FFT-tp3694587p3694589.html

High Andrew,

make sure that you set the pixels for masking to exactly zero.
If you set them to, e.g. 1, you can't see a difference, but it won't  
do anything.

Also I'd suggest that you upgrade to 1.41k or later: Then it is  
enough if you mask the stripes on only one side; ImageJ will add the  
mask on the other side for you (with better symmetry than you could  
get manually).

For your application, you might also try the FFT>Bandpass filter with  
"filter small structures" set to 0, "filter large structures" at a  
very high value (e.g. a million), "suppress stripes" set to  
"horizontal" and a percentage value that you have to find by trial  
and error.

Michael
________________________________________________________________

On 31 Oct 2008, at 16:42, Andrew Vaughan wrote:

> Dear All.
>
> I've been trying to remove some line frequency interference from  
> confocal images by generating FFTs
> of the images and masking the frequencies using the region  
> selection and fill method mentioned in
> the online documentation. However, when I reverse Fourier transform  
> the frequency domain image
> the lines are still there. When I subsequently perform another FFT  
> on the resulting image I find that
> the region(s) have been ignored and the frequency domain image  
> appears identical to the original.
> I'm using Image J version 1.41g on an Intel Mac running OS X  
> 10.4.11. The images are either slices
> from Bio-Rad pic files or TIFFs of these.
>
> Thanks.
>
> Yours sincerely,
>
> Andrew Vaughan.