Non-uniform illumination and noise reduction

[MikeChelen]

Hello,
I'm trying to analyze some calcium dye fluorescence images. The
illuminating field appears to be non-uniform, and I'm looking for a
good way to calculate and remove the background illumination. So far,
I've tried both the built-in "Subtract Background" command (Process ->
Subtract Background) and the plugin "Background Correction"
(http://rsb.info.nih.gov/ij/plugins/background.html). These both to
seem to work reasonably well, but the second does not work for a
multiframe image. I was wondering if anyone had any experience with
these two tools, and if they could provide any insight into how these
tools work, and if there are others I should be considering.

Because the image set is a 500 frame t-series, it seems like the best
algorithm would be one that uses the data from every frame to
calculate the best possible approximation of the illuminating field.
Additionally, such a utility could identify likely points of noise
because the noise will change much more rapidly (from frame to frame)
than the actual signal. Is there anything like this for ImageJ that
anyone can recommend? I would really appreciate any suggestions or
insight.

Thanks,
Michael Chelen

[MikeChelen]

Hello,
I'm trying to analyze some calcium dye fluorescence images. The
illuminating field appears to be non-uniform, and I'm looking for a
good way to calculate and remove the background illumination. So far,
I've tried both the built-in "Subtract Background" command (Process ->
Subtract Background) and the plugin "Background Correction"
(http://rsb.info.nih.gov/ij/plugins/background.html). These both to
seem to work reasonably well, but the second does not work for a
multiframe image. I was wondering if anyone had any experience with
these two tools, and if they could provide any insight into how these
tools work, and if there are others I should be considering.

Because the image set is a 500 frame t-series, it seems like the best
algorithm would be one that uses the data from every frame to
calculate the best possible approximation of the illuminating field.
Additionally, such a utility could identify likely points of noise
because the noise will change much more rapidly (from frame to frame)
than the actual signal. Is there anything like this for ImageJ that
anyone can recommend? I would really appreciate any suggestions or
insight.

Thanks,
Michael Chelen

[Joel Sheffield]

Can you get an image of the illuminating field without the sample?  
If so, you could subtract, or divide (depending on what works for
you)this image from each of the frames, and then adjust b/c.  
Alternatively, you might be able to create a pseudo background by
taking one of your original images and carrying out a blur operation
to reduce the effect of the sample, but retain the broad intensity
distribution.

Joel

> Hello,
> I'm trying to analyze some calcium dye fluorescence images. The
> illuminating field appears to be non-uniform, and I'm looking for a
> good way to calculate and remove the background illumination. So far,
> I've tried both the built-in "Subtract Background" command (Process ->
> Subtract Background) and the plugin "Background Correction"
> (http://rsb.info.nih.gov/ij/plugins/background.html). These both to
> seem to work reasonably well, but the second does not work for a
> multiframe image. I was wondering if anyone had any experience with
> these two tools, and if they could provide any insight into how these
> tools work, and if there are others I should be considering.
>
> Because the image set is a 500 frame t-series, it seems like the best
> algorithm would be one that uses the data from every frame to
> calculate the best possible approximation of the illuminating field.
> Additionally, such a utility could identify likely points of noise
> because the noise will change much more rapidly (from frame to frame)
> than the actual signal. Is there anything like this for ImageJ that
> anyone can recommend? I would really appreciate any suggestions or
> insight.
>
> Thanks,
> Michael Chelen

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

[Henry Barwood]

I have found that taking a "blank" image (something like just a glass slide)
and subtracting this from the images works really well to even the
illumination. You have to adjust the blank using image math to get the right
degree of subtraction (too light and the resulting image will be dark, too
dark and you will have a very light result). Of course, this is time
consuming and I have not worked with a stack.

Henry Barwood

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]]On Behalf Of
Michael Chelen
Sent: Thursday, August 25, 2005 9:30 AM
To: [hidden email]
Subject: Non-uniform illumination and noise reduction


Hello,
I'm trying to analyze some calcium dye fluorescence images. The
illuminating field appears to be non-uniform, and I'm looking for a
good way to calculate and remove the background illumination. So far,
I've tried both the built-in "Subtract Background" command (Process ->
Subtract Background) and the plugin "Background Correction"
(http://rsb.info.nih.gov/ij/plugins/background.html). These both to
seem to work reasonably well, but the second does not work for a
multiframe image. I was wondering if anyone had any experience with
these two tools, and if they could provide any insight into how these
tools work, and if there are others I should be considering.

Because the image set is a 500 frame t-series, it seems like the best
algorithm would be one that uses the data from every frame to
calculate the best possible approximation of the illuminating field.
Additionally, such a utility could identify likely points of noise
because the noise will change much more rapidly (from frame to frame)
than the actual signal. Is there anything like this for ImageJ that
anyone can recommend? I would really appreciate any suggestions or
insight.

Thanks,
Michael Chelen

[R Wade Schuette]

I'm not sure what your images look like, but you might be able to
compute a minimum or maximum pixel-value across the stack for each
pixel, then truncate or smooth that to get a "standard background".

Wade


On Thu, 25 Aug 2005, Michael Chelen wrote:

> Hello,
> I'm trying to analyze some calcium dye fluorescence images. The
> illuminating field appears to be non-uniform, and I'm looking for a
> good way to calculate and remove the background illumination. So far,
> I've tried both the built-in "Subtract Background" command (Process ->
> Subtract Background) and the plugin "Background Correction"
> (http://rsb.info.nih.gov/ij/plugins/background.html). These both to
> seem to work reasonably well, but the second does not work for a
> multiframe image. I was wondering if anyone had any experience with
> these two tools, and if they could provide any insight into how these
> tools work, and if there are others I should be considering.
>
> Because the image set is a 500 frame t-series, it seems like the best
> algorithm would be one that uses the data from every frame to
> calculate the best possible approximation of the illuminating field.
> Additionally, such a utility could identify likely points of noise
> because the noise will change much more rapidly (from frame to frame)
> than the actual signal. Is there anything like this for ImageJ that
> anyone can recommend? I would really appreciate any suggestions or
> insight.
>
> Thanks,
> Michael Chelen
>
>
>

[Michael Elbaum]

I second the suggestion. Most likely the non-uniformity in your
illumination field does not change or flicker in time. (if it does you
should probably replace the Hg lamp!) You can take a blank image simply
with some free fluorophore in solution. Then you need to boost
sensitivity in the dim regions or suppress it in the bright ones,
relatively. This is "flat-field"ing with what ought to be a uniformly
bright image. You should divide each frame by the blank, with some
appropriate normalization in the arithmetic to match the brightnesses.
If you want to keep things quantitative you should first subtract a dark
image from your stack, or the average value of a dark image, so that
zero light gives zero pixel value. You should subtract it from the blank
as well, but make sure there are no zero values. In the end each pixel
should have the same sensitivity to fluorescence with no dependence on
its position.

good luck...
Michael Elbaum



>>> Joel Sheffield <[hidden email]> 08/25/05 6:36 PM >>>
Can you get an image of the illuminating field without the sample?  
If so, you could subtract, or divide (depending on what works for
you)this image from each of the frames, and then adjust b/c.  
Alternatively, you might be able to create a pseudo background by
taking one of your original images and carrying out a blur operation
to reduce the effect of the sample, but retain the broad intensity
distribution.

Joel

> Hello,
> I'm trying to analyze some calcium dye fluorescence images. The
> illuminating field appears to be non-uniform, and I'm looking for a
> good way to calculate and remove the background illumination. So far,
> I've tried both the built-in "Subtract Background" command (Process ->
> Subtract Background) and the plugin "Background Correction"
> (http://rsb.info.nih.gov/ij/plugins/background.html). These both to
> seem to work reasonably well, but the second does not work for a
> multiframe image. I was wondering if anyone had any experience with
> these two tools, and if they could provide any insight into how these
> tools work, and if there are others I should be considering.
>
> Because the image set is a 500 frame t-series, it seems like the best
> algorithm would be one that uses the data from every frame to
> calculate the best possible approximation of the illuminating field.
> Additionally, such a utility could identify likely points of noise
> because the noise will change much more rapidly (from frame to frame)
> than the actual signal. Is there anything like this for ImageJ that
> anyone can recommend? I would really appreciate any suggestions or
> insight.
>
> Thanks,
> Michael Chelen

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

[Smith, Mike]

Michael,

I have no way of knowing what your tolerance is for non-uniform illumination, but are your scope and fluorescence system properly aligned? A research grade microscope should have reasonably uniform illumination  when properly set up. Apologies if this seems simplistic, but I have found this is often the cause of illumination problems.

Mike Smith


-----Original Message-----
From: ImageJ Interest Group on behalf of Michael Chelen
Sent: Thu 8/25/2005 10:29 AM
To: [hidden email]
Subject: Non-uniform illumination and noise reduction
 
Hello,
I'm trying to analyze some calcium dye fluorescence images. The
illuminating field appears to be non-uniform, and I'm looking for a
good way to calculate and remove the background illumination. So far,
I've tried both the built-in "Subtract Background" command (Process ->
Subtract Background) and the plugin "Background Correction"
(http://rsb.info.nih.gov/ij/plugins/background.html). These both to
seem to work reasonably well, but the second does not work for a
multiframe image. I was wondering if anyone had any experience with
these two tools, and if they could provide any insight into how these
tools work, and if there are others I should be considering.

Because the image set is a 500 frame t-series, it seems like the best
algorithm would be one that uses the data from every frame to
calculate the best possible approximation of the illuminating field.
Additionally, such a utility could identify likely points of noise
because the noise will change much more rapidly (from frame to frame)
than the actual signal. Is there anything like this for ImageJ that
anyone can recommend? I would really appreciate any suggestions or
insight.

Thanks,
Michael Chelen

[Robert Dougherty]

A way to compute the background might be to try Polynomial Fit.  It
seems like this is a pluin in search of an application.
http://www.optinav.com/ImageJplugins/Polynomial_Fit.htm

Robert P. Dougherty, Ph.D.
President, OptiNav, Inc.
[phone]
[hidden email]

>
>
> -----Original Message-----
> From: ImageJ Interest Group on behalf of Michael Chelen
> Sent: Thu 8/25/2005 10:29 AM
> To: [hidden email]
> Subject: Non-uniform illumination and noise reduction
>
> Hello,
> I'm trying to analyze some calcium dye fluorescence images. The
> illuminating field appears to be non-uniform, and I'm looking for a
> good way to calculate and remove the background illumination. So far,
> I've tried both the built-in "Subtract Background" command (Process ->
> Subtract Background) and the plugin "Background Correction"
> (http://rsb.info.nih.gov/ij/plugins/background.html). These both to
> seem to work reasonably well, but the second does not work for a
> multiframe image. I was wondering if anyone had any experience with
> these two tools, and if they could provide any insight into how these
> tools work, and if there are others I should be considering.
>
> Because the image set is a 500 frame t-series, it seems like the best
> algorithm would be one that uses the data from every frame to
> calculate the best possible approximation of the illuminating field.
> Additionally, such a utility could identify likely points of noise
> because the noise will change much more rapidly (from frame to frame)
> than the actual signal. Is there anything like this for ImageJ that
> anyone can recommend? I would really appreciate any suggestions or
> insight.
>
> Thanks,
> Michael Chelen

[Albert Cardona]

I find the Calculator Plus plugin most useful for the purpose of
eliminating noise. Just follow the instructions on the Wright ImageJ
manual at

http://www.uhnresearch.ca/facilities/wcif/download.php

The process can be automated from a macro to process each slice of a
stack.

Albert


Albert Cardona
Institute of Neuroinformatics     Tel : [phone]
University/ETH Zurich             Fax : [phone]
Winterthurerstrasse 190           acardona (at) ini phys ethz ch
Zurich 8057, Switzerland          www.ini.unizh.ch