FFT-bandpass filter: exact meaning of low- and high-pass thresshold

[Jan-Peter Urbach]

Hello,
I am just starting to use ImageJ for Image analysis and I am really
impressed by the various functionalities.
Particularily I want to convolute images with a defined convolution kernel.
This is where I have my question:
To my understanding a convolution with a gaussian kernel can be
accomplished by the Process -> FFT -> Bandpass Filter menu.
There you can input the thresholds for large and small structures to be
filtered out.
My question: What exactly do those number mean? Or more precisly the
function effectivly performs a convolution of the image in real space with
a gaussian kernel of a given size (well actually two kernels one for the
large and one for the small structures).
What is the relationship between the threshold you enter in the menu and
the size of the gaussian kernel in real space.
For example is it the FWHM (full width half max) or is it the standard
deviation of a normalized gaussian distribution?

Thanks in advance for any hints
Jan-Peter

[H. Gluender]

>Hello,
>I am just starting to use ImageJ for Image analysis and I am really
>impressed by the various functionalities.
>Particularily I want to convolute images with a defined convolution kernel.
>This is where I have my question:
>To my understanding a convolution with a gaussian kernel can be
>accomplished by the Process -> FFT -> Bandpass Filter menu.
>There you can input the thresholds for large and small structures to be
>filtered out.
>My question: What exactly do those number mean? Or more precisly the
>function effectivly performs a convolution of the image in real space with
>a gaussian kernel of a given size (well actually two kernels one for the
>large and one for the small structures).
>What is the relationship between the threshold you enter in the menu and
>the size of the gaussian kernel in real space.
>For example is it the FWHM (full width half max) or is it the standard
>deviation of a normalized gaussian distribution?
>
>Thanks in advance for any hints
>Jan-Peter

Did you look at

<http://rsb.info.nih.gov/ij/plugins/fft-filter.html>

as recommended in the ImageJ-manual?

Best
--


                   Herbie

          ------------------------

          <http://www.gluender.de>

[Jan-Peter Urbach]

Yes, I did.
The only hint about the size of the filtering :
...
Filters out large structures (shading correction) and small structures
(smoothing) of the specified size by gaussian filtering in fourier
space.
...
does not seem to answer my question or am I missing something?
JP



---------------------------------------------------------------------
Dr. Jan-Peter Urbach
QAG PDT CS CADB
Qimonda AG
Phone: [phone]
e-mail: [hidden email]
Visitor Address:  Am Campeon 1 - 12, D-85579 Neubiberg, Room 12.3.101
        *** visit our homepage at http://www.qimonda.com ***

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of H.
Gluender
Sent: Tuesday, September 05, 2006 4:16 PM
To: [hidden email]
Subject: Re: FFT-bandpass filter: exact meaning of low- and high-pass
thresshold

>Hello,
>I am just starting to use ImageJ for Image analysis and I am really
>impressed by the various functionalities.
>Particularily I want to convolute images with a defined convolution
kernel.
>This is where I have my question:
>To my understanding a convolution with a gaussian kernel can be
>accomplished by the Process -> FFT -> Bandpass Filter menu.
>There you can input the thresholds for large and small structures to be

>filtered out.
>My question: What exactly do those number mean? Or more precisly the
>function effectivly performs a convolution of the image in real space
>with a gaussian kernel of a given size (well actually two kernels one
>for the large and one for the small structures).
>What is the relationship between the threshold you enter in the menu
>and the size of the gaussian kernel in real space.
>For example is it the FWHM (full width half max) or is it the standard
>deviation of a normalized gaussian distribution?
>
>Thanks in advance for any hints
>Jan-Peter

Did you look at

<http://rsb.info.nih.gov/ij/plugins/fft-filter.html>

as recommended in the ImageJ-manual?

Best
--


                   Herbie

          ------------------------

          <http://www.gluender.de>

[Joel Sheffield]

You can get a detailed image of the filter mask by checking the
appropriate box in the menu.  That should help.

Joel

> Yes, I did.
> The only hint about the size of the filtering :
> ...
> Filters out large structures (shading correction) and small structures
> (smoothing) of the specified size by gaussian filtering in fourier
> space. ... does not seem to answer my question or am I missing
> something? JP
>
>
>
> ---------------------------------------------------------------------
> Dr. Jan-Peter Urbach QAG PDT CS CADB Qimonda AG Phone: [phone]
> 28645 e-mail: [hidden email] Visitor Address:  Am
> Campeon 1 - 12, D-85579 Neubiberg, Room 12.3.101
>  *** visit our homepage at http://www.qimonda.com ***
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
> H. Gluender Sent: Tuesday, September 05, 2006 4:16 PM To:
> [hidden email] Subject: Re: FFT-bandpass filter: exact meaning of
> low- and high-pass thresshold
>
> >Hello,
> >I am just starting to use ImageJ for Image analysis and I am really
> >impressed by the various functionalities. Particularily I want to
> >convolute images with a defined convolution
> kernel.
> >This is where I have my question:
> >To my understanding a convolution with a gaussian kernel can be
> >accomplished by the Process -> FFT -> Bandpass Filter menu. There you
> >can input the thresholds for large and small structures to be
>
> >filtered out.
> >My question: What exactly do those number mean? Or more precisly the
> >function effectivly performs a convolution of the image in real space
> > with a gaussian kernel of a given size (well actually two kernels
> >one for the large and one for the small structures). What is the
> >relationship between the threshold you enter in the menu and the size
> >of the gaussian kernel in real space. For example is it the FWHM
> >(full width half max) or is it the standard deviation of a normalized
> >gaussian distribution?
> >
> >Thanks in advance for any hints
> >Jan-Peter
>
> Did you look at
>
> <http://rsb.info.nih.gov/ij/plugins/fft-filter.html>
>
> as recommended in the ImageJ-manual?
>
> Best
> --
>
>
>                    Herbie
>
>           ------------------------
>
>           <http://www.gluender.de>

--
Joel B. Sheffield, Ph.D.
Biology Department, Temple University
1900 North 12th Street
Philadelphia, PA 19122
[hidden email]  
[phone], fax [phone]
http://astro.temple.edu/~jbs