Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

>
> you are correct, but I want to pose some caution. This technique can be
> effective in removing periodic artifacts but one has to be aware that by
> cutting out portions of the spectrum you also remove these portions from the
> spectrum of your image data. Further, it is advised to use some continuous
> function that is similar in reciprocal space (e.g. gauss, cosine...) to
> weight your cutting window. This will supress oscillatory artifacts around
> objects after the inverse transform. Finally, performing that filtering in a
> larger, mean-apodized window (with borders of half the size of your filter
> window in reciprocal space) will eliminate effects on the edges that are
> imposed by the discontinuity of the Filter and arises due to the circulartiy
> of the Fourier transform.
>

Like any filtering method, it works when you can use
a characterization that separates the noise from the desired information.
In this case, you need for the spatial frequency of the disturbance to be
different from those of the image. This is often the case, especially for
DIC images. While it is always better to solve the underlying problem,
I have used this method to save data sets that were already acquired and
could not be repeated.

As with all filtering methods, it is not always appropriate. However, it is
one of a number of very useful methods that is readily available and
often overlooked.

--aryeh
Aryeh Weiss                          | email: [log in to unmask]
Department of Electronics            | URL:   http://optics.jct.ac.il/~aryeh
Jerusalem College of Technology      | phone: 972-2-6751146
POB 16031                            | FAX:   972-2-6751275
Jerusalem, Israel                    | ham radio: 4X1PB/KA1PB