William,
still not sure where the problem is.
You mention that you've studied the relevant sections of the ImageJ user
guide (pages 120 to 123).
From this text it is obvious that there are three approaches to spatial
frequency filtering of images. They require more or less work on your
side but they will also produce results of (considerably) different quality.
1. The easiest and quickest approach is described in section 29.10.2.
The drawback is that you cannot influence the gray-value shape of the
mask, i.e. the gray-value slopes of the blocking dots or areas. In fact
you draw a binary mask that actually is somehow blurred automatically to
avoid the hard borders of the mask which reduces artifacts, but you will
not be able to fine tune the gray-value slopes. A great advantage is
that the filter function (mask) is automatically made symmetric to the
origin, i.e. you obtain the correct filter function from specifying a
single sideband function. (It is assumed that single sideband filtering
is not desired!)
2. A better (quality) approach is to use "custom filters" described in
section 29.10.6.
With this approach you can nicely define the desired filter functions as
non-binary (gray-value) images. However, you have to take care for the
mentioned symmetry, and in general you need to analyze the spectra of
your images to design the filter functions properly.
3. You do it all "by hand".
This approach allows you to use 32-bit filter functions. Do a FFT of
your image, multiply it by a properly designed filter function, and
re-transform the product.
In any case, be aware that you need square sized images of side lengths
that are a power of two.
HTH
Herbie
::::::::::::::::::::::::::::::::::::::::
On 17.04.14 23:31, Clark, William wrote:
> Herbie
>
> OK, “atomic resolution image”, then. I am filtering it to enhance
> the atom columns, and I am simply not familiar with the process in
> ImageJ for creating the mask and then adding it to the FFT (which I
> already have) before inverse transforming back to the filtered image.
> I have looked in the manual and on the web, but I have not found what
> I need. I have never used ImageJ for doing this - I have used other
> software packages to do this in the past.
>
> Cheers,
>
> William
>
>
> On Apr 17, 2014, at 2:50 PM🌛, Herbie
> <
[hidden email]<mailto:
[hidden email]>> wrote:
>
> William,
>
> could you please be more specific?
>
> Why do you think "high resolution" may be somehow special?
>
> Where exactly is your problem with creating filter functions?
>
> What do you expect from the filtering process? Enhanced
> periodicities?
>
> Did you already have a look at Fourier spectra of your samples?
>
> Regards
>
> Herbie
>
> ::::::::::::::::::::::::::::::::::::::: On 17.04.14 18:23, William
> Clark wrote: Does anyone have a description of the ImageJ procedures
> for Fourier filtering a high resolution (S)TEM image of a crystalline
> sample? In particular I would like to know the best way to create and
> apply the filtering mask to the FFT, and then invert them.
>
> Any help gratefully received!
>
> -- ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>
> -- ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>
>
> [cid:6B93ECAC-77A8-4FEE-AFC4-63C4C3E946B2]
>
> William A. T. Clark, D. Phil., C. Eng., FASM Professor Department of
> Materials Science & Engineering 480 Watts Hall, 2041 College Road,
> COLUMBUS, OH 43210 614-292-0575 Office: 614-747-6266 Mobile:
> 614-292-1537 Fax
[hidden email]<mailto:
[hidden email]>
>
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html
Locked
