Two channel color merge

[Knecht, David]

In a red-green merge of two channel data, I find that red stands out well from green and yellow, but green and yellow are often very difficult to distinguish from each other (to my eye).  Without getting into all the issues of measuring colocalization, is there a better color combination or two channel LUT to be able to visualize each channel and overlap clearly?  
Also, what is the best way to try to balance the intensities of the two images before merging.  If you have an area that you want to target as "yellow" because it looks like an area of colocalization, and that area in the two images have different pixel intensities, how do you get them matched? When you are using the B/C panel, you can't see what the intensity will be after the LUT is applied.  It might be nice to have ImageJ be aable to read out those values before applying so you can do this more accurately.  Perhaps a second set of values at the top of the image that reflect the values if the current LUT were applied?  Dave

 
Dr. David Knecht    
Department of Molecular and Cell Biology
Co-head Flow Cytometry and Confocal Microscopy Facility
U-3125
91 N. Eagleville Rd.
University of Connecticut
Storrs, CT 06269
860-486-2200
860-486-4331 (fax)

[Joel Sheffield]

David,

You are describing a problem that about 10% of use deuteranopes have to
variable degrees.  One solution is to double the red channel with blue,
creating magenta.  In such a 3 color image, the green channel is , well,
green.  The red channel is magenta, and overlaps are white.  Naturally, as
you suggest, this works best if the intensity ranges of each are about the
same.

Some time ago, with Wayne's help, I wrote a macro that does this conversion:

Here 'tis:

// Recolor.txt
// This macro converts an rgb image to one in which
// the red channel is replaced by magenta to allow
// easier viewing of images with multiple fluorescence
// by color blind viewers.
//
// Author: Joel Sheffield, Biology Department, Temple University
// 2/22/05 - now works with stacks and single images.

   requires("1.33o");
   setBatchMode(true);
   orig = getTitle();
   run("Duplicate...", "title=Copy duplicate");
   run("RGB Split");
   selectImage("Copy (blue)");
   run("Blue");
   run("RGB Color");
   run("Rename...", "title=blue");
   selectImage("Copy (green)");
   run("Green");
   run("RGB Color");
   run("Rename...", "title=green");
   selectImage("Copy (red)");
   run("Magenta");
   run("RGB Color");
   run("Rename...", "title=red");
   run("Image Calculator...", "image1=red operation=Add image2=blue create
stack");
   run("Image Calculator...", "image1='Result of red' operation=Add
image2=green stack");
   newtitle= "Recolor-" +orig;
   run("Rename...", "title=[" +newtitle+"]");
   selectImage("red");
   close();
   selectImage("blue");
   close();
   selectImage("green");
   close();
   selectImage(newtitle);
   setBatchMode(false);

Best of luck.

Joel




On Mon, Jul 19, 2010 at 2:25 PM, David Knecht <[hidden email]>wrote:

> In a red-green merge of two channel data, I find that red stands out well
> from green and yellow, but green and yellow are often very difficult to
> distinguish from each other (to my eye).  Without getting into all the
> issues of measuring colocalization, is there a better color combination or
> two channel LUT to be able to visualize each channel and overlap clearly?
> Also, what is the best way to try to balance the intensities of the two
> images before merging.  If you have an area that you want to target as
> "yellow" because it looks like an area of colocalization, and that area in
> the two images have different pixel intensities, how do you get them
> matched? When you are using the B/C panel, you can't see what the intensity
> will be after the LUT is applied.  It might be nice to have ImageJ be aable
> to read out those values before applying so you can do this more accurately.
>  Perhaps a second set of values at the top of the image that reflect the
> values if the current LUT were applied?  Dave
>
>
> Dr. David Knecht
> Department of Molecular and Cell Biology
> Co-head Flow Cytometry and Confocal Microscopy Facility
> U-3125
> 91 N. Eagleville Rd.
> University of Connecticut
> Storrs, CT 06269
> 860-486-2200
> 860-486-4331 (fax)
>

--


Joel B. Sheffield, Ph.D
Department of Biology
Temple University
Philadelphia, PA 19122
Voice: 215 204 8839
e-mail: [hidden email]
URL:  http://astro.temple.edu/~jbs

[ctrueden]

Hi David,

I'm not sure this addresses every aspect of your question, but have you
tried the composite image functionality? Choose Image > Color > Make
Composite from the menu. You can then quickly control the LUT for each
channel separately using the Channels dialog that appears. Move the C slider
to the desired channel, hit the "More >>" button and select the desired LUT.
There are a few presets, or you can load a custom one using the "Edit
LUT..." option. In this way you can play with composite color combinations
to find the one that looks best for you.

-Curtis

On Mon, Jul 19, 2010 at 1:25 PM, David Knecht <[hidden email]>wrote:

> In a red-green merge of two channel data, I find that red stands out well
> from green and yellow, but green and yellow are often very difficult to
> distinguish from each other (to my eye).  Without getting into all the
> issues of measuring colocalization, is there a better color combination or
> two channel LUT to be able to visualize each channel and overlap clearly?
> Also, what is the best way to try to balance the intensities of the two
> images before merging.  If you have an area that you want to target as
> "yellow" because it looks like an area of colocalization, and that area in
> the two images have different pixel intensities, how do you get them
> matched? When you are using the B/C panel, you can't see what the intensity
> will be after the LUT is applied.  It might be nice to have ImageJ be aable
> to read out those values before applying so you can do this more accurately.
>  Perhaps a second set of values at the top of the image that reflect the
> values if the current LUT were applied?  Dave
>
>
> Dr. David Knecht
> Department of Molecular and Cell Biology
> Co-head Flow Cytometry and Confocal Microscopy Facility
> U-3125
> 91 N. Eagleville Rd.
> University of Connecticut
> Storrs, CT 06269
> 860-486-2200
> 860-486-4331 (fax)
>

[Knecht, David]

Hi Joel- You can do that type of merge with the Image/Color/Merge Channels and select the red channel for blue and red.  Does the macro do something different?  I tried it on a test image and It is an interesting effect.  I will play with it and see how it works.  Thanks- Dave

On Jul 19, 2010, at 2:58 PM, JOEL B. SHEFFIELD wrote:

> David,
>
> You are describing a problem that about 10% of use deuteranopes have to
> variable degrees.  One solution is to double the red channel with blue,
> creating magenta.  In such a 3 color image, the green channel is , well,
> green.  The red channel is magenta, and overlaps are white.  Naturally, as
> you suggest, this works best if the intensity ranges of each are about the
> same.
>
> Some time ago, with Wayne's help, I wrote a macro that does this conversion:
>
> Here 'tis:
>
> // Recolor.txt
> // This macro converts an rgb image to one in which
> // the red channel is replaced by magenta to allow
> // easier viewing of images with multiple fluorescence
> // by color blind viewers.
> //
> // Author: Joel Sheffield, Biology Department, Temple University
> // 2/22/05 - now works with stacks and single images.
>
>   requires("1.33o");
>   setBatchMode(true);
>   orig = getTitle();
>   run("Duplicate...", "title=Copy duplicate");
>   run("RGB Split");
>   selectImage("Copy (blue)");
>   run("Blue");
>   run("RGB Color");
>   run("Rename...", "title=blue");
>   selectImage("Copy (green)");
>   run("Green");
>   run("RGB Color");
>   run("Rename...", "title=green");
>   selectImage("Copy (red)");
>   run("Magenta");
>   run("RGB Color");
>   run("Rename...", "title=red");
>   run("Image Calculator...", "image1=red operation=Add image2=blue create
> stack");
>   run("Image Calculator...", "image1='Result of red' operation=Add
> image2=green stack");
>   newtitle= "Recolor-" +orig;
>   run("Rename...", "title=[" +newtitle+"]");
>   selectImage("red");
>   close();
>   selectImage("blue");
>   close();
>   selectImage("green");
>   close();
>   selectImage(newtitle);
>   setBatchMode(false);
>
> Best of luck.
>
> Joel
>
>
>
>
> On Mon, Jul 19, 2010 at 2:25 PM, David Knecht <[hidden email]>wrote:
>
>> In a red-green merge of two channel data, I find that red stands out well
>> from green and yellow, but green and yellow are often very difficult to
>> distinguish from each other (to my eye).  Without getting into all the
>> issues of measuring colocalization, is there a better color combination or
>> two channel LUT to be able to visualize each channel and overlap clearly?
>> Also, what is the best way to try to balance the intensities of the two
>> images before merging.  If you have an area that you want to target as
>> "yellow" because it looks like an area of colocalization, and that area in
>> the two images have different pixel intensities, how do you get them
>> matched? When you are using the B/C panel, you can't see what the intensity
>> will be after the LUT is applied.  It might be nice to have ImageJ be aable
>> to read out those values before applying so you can do this more accurately.
>> Perhaps a second set of values at the top of the image that reflect the
>> values if the current LUT were applied?  Dave
>>
>>
>> Dr. David Knecht
>> Department of Molecular and Cell Biology
>> Co-head Flow Cytometry and Confocal Microscopy Facility
>> U-3125
>> 91 N. Eagleville Rd.
>> University of Connecticut
>> Storrs, CT 06269
>> 860-486-2200
>> 860-486-4331 (fax)
>>
>
>
>
> --
>
>
> Joel B. Sheffield, Ph.D
> Department of Biology
> Temple University
> Philadelphia, PA 19122
> Voice: 215 204 8839
> e-mail: [hidden email]
> URL:  http://astro.temple.edu/~jbs

Dr. David Knecht    
Department of Molecular and Cell Biology
Co-head Flow Cytometry and Confocal Microscopy Facility
U-3125
91 N. Eagleville Rd.
University of Connecticut
Storrs, CT 06269
860-486-2200
860-486-4331 (fax)

[Joel Sheffield]

Hi Dave,

You are right, of course.  I developed this in order to look at published
pictures, or student's theses (!)

Joel

On Mon, Jul 19, 2010 at 4:47 PM, David Knecht <[hidden email]>wrote:

> Hi Joel- You can do that type of merge with the Image/Color/Merge Channels
> and select the red channel for blue and red.  Does the macro do something
> different?  I tried it on a test image and It is an interesting effect.  I
> will play with it and see how it works.  Thanks- Dave
>
> On Jul 19, 2010, at 2:58 PM, JOEL B. SHEFFIELD wrote:
>
> > David,
> >
> > You are describing a problem that about 10% of use deuteranopes have to
> > variable degrees.  One solution is to double the red channel with blue,
> > creating magenta.  In such a 3 color image, the green channel is , well,
> > green.  The red channel is magenta, and overlaps are white.  Naturally,
> as
> > you suggest, this works best if the intensity ranges of each are about
> the
> > same.
> >
> > Some time ago, with Wayne's help, I wrote a macro that does this
> conversion:
> >
> > Here 'tis:
> >
> > // Recolor.txt
> > // This macro converts an rgb image to one in which
> > // the red channel is replaced by magenta to allow
> > // easier viewing of images with multiple fluorescence
> > // by color blind viewers.
> > //
> > // Author: Joel Sheffield, Biology Department, Temple University
> > // 2/22/05 - now works with stacks and single images.
> >
> >   requires("1.33o");
> >   setBatchMode(true);
> >   orig = getTitle();
> >   run("Duplicate...", "title=Copy duplicate");
> >   run("RGB Split");
> >   selectImage("Copy (blue)");
> >   run("Blue");
> >   run("RGB Color");
> >   run("Rename...", "title=blue");
> >   selectImage("Copy (green)");
> >   run("Green");
> >   run("RGB Color");
> >   run("Rename...", "title=green");
> >   selectImage("Copy (red)");
> >   run("Magenta");
> >   run("RGB Color");
> >   run("Rename...", "title=red");
> >   run("Image Calculator...", "image1=red operation=Add image2=blue create
> > stack");
> >   run("Image Calculator...", "image1='Result of red' operation=Add
> > image2=green stack");
> >   newtitle= "Recolor-" +orig;
> >   run("Rename...", "title=[" +newtitle+"]");
> >   selectImage("red");
> >   close();
> >   selectImage("blue");
> >   close();
> >   selectImage("green");
> >   close();
> >   selectImage(newtitle);
> >   setBatchMode(false);
> >
> > Best of luck.
> >
> > Joel
> >
> >
> >
> >
> > On Mon, Jul 19, 2010 at 2:25 PM, David Knecht <[hidden email]
> >wrote:
> >
> >> In a red-green merge of two channel data, I find that red stands out
> well
> >> from green and yellow, but green and yellow are often very difficult to
> >> distinguish from each other (to my eye).  Without getting into all the
> >> issues of measuring colocalization, is there a better color combination
> or
> >> two channel LUT to be able to visualize each channel and overlap
> clearly?
> >> Also, what is the best way to try to balance the intensities of the two
> >> images before merging.  If you have an area that you want to target as
> >> "yellow" because it looks like an area of colocalization, and that area
> in
> >> the two images have different pixel intensities, how do you get them
> >> matched? When you are using the B/C panel, you can't see what the
> intensity
> >> will be after the LUT is applied.  It might be nice to have ImageJ be
> aable
> >> to read out those values before applying so you can do this more
> accurately.
> >> Perhaps a second set of values at the top of the image that reflect the
> >> values if the current LUT were applied?  Dave
> >>
> >>
> >> Dr. David Knecht
> >> Department of Molecular and Cell Biology
> >> Co-head Flow Cytometry and Confocal Microscopy Facility
> >> U-3125
> >> 91 N. Eagleville Rd.
> >> University of Connecticut
> >> Storrs, CT 06269
> >> 860-486-2200
> >> 860-486-4331 (fax)
> >>
> >
> >
> >
> > --
> >
> >
> > Joel B. Sheffield, Ph.D
> > Department of Biology
> > Temple University
> > Philadelphia, PA 19122
> > Voice: 215 204 8839
> > e-mail: [hidden email]
> > URL:  http://astro.temple.edu/~jbs <http://astro.temple.edu/%7Ejbs>
>
> Dr. David Knecht
> Department of Molecular and Cell Biology
> Co-head Flow Cytometry and Confocal Microscopy Facility
> U-3125
> 91 N. Eagleville Rd.
> University of Connecticut
> Storrs, CT 06269
> 860-486-2200
> 860-486-4331 (fax)
>

--


Joel B. Sheffield, Ph.D
Department of Biology
Temple University
Philadelphia, PA 19122
Voice: 215 204 8839
e-mail: [hidden email]
URL:  http://astro.temple.edu/~jbs

[Gabriel Landini]

On Monday 19 July 2010 21:47:40 you wrote:
> Hi Joel- You can do that type of merge with the Image/Color/Merge Channels
> and select the red channel for blue and red.  Does the macro do something
> different?  I tried it on a test image and It is an interesting effect.  I
> will play with it and see how it works.

Hi,
It is close but not quite the same as using the red for R and for B.
 
In the method that Joel describes, in the blue channel one puts the result of
Red+Blue and the red channel remains as is. So a trichromatic individual would
still see 3 channels/colours as far as they do not colocalise (green, magenta
(r+b), and blue).
However the method is best for data from 2 channels so both dichromats and
trichromats can see the same differences (although in differently perceived
colours).

Incidentally at the Luxembourg conference I hope to be presenting some
techniques for image recolouring for dichromatic observers (additive images,
subtractive images, look up tables and simulations of colour blindness).

Cheers

Gabriel

[Diamond]

I am jumping on the thread a tad late, but i have similar problems and hope to get some help. I have two channels in whose colocalization i am interested in. Both the channel are of different intensities and and hence i am having a trouble getting a "yellow" merge image.

From literature available, i understand that the following procedure may serve better to show colocalization in my case. I am not sure whether i have understood it right and i would appreciate your feedback.

1) Apply Magenta LUT to my red channel.
2) Apply Cyan LUT to my green channel.
3) Covert the images to RGB and perform a difference operation using Image Calculator or Use color function -color merge using the difference operator.
4) In the resulting image: Yellow reflects colocalization where the pixels are of equal intensity.
                                    Green reflects colocalization with cyan being of brighter intensity than magenta.
                                    Red reflects colocalization with magenta being of brighter intensity than cyan.

I have performed quantification on my images and it does suggest colocalization. Is the above mentioned method correct if i want to publish my images ?

Would really really appreciate your response !
Thanks a lot,
Cheers
Heera

[dscho]

Dear Diamond,

On Sun, 30 Oct 2011, Diamond wrote:

> I am jumping on the thread a tad late, but i have similar problems and
> hope to get some help. I have two channels in whose colocalization i am
> interested in. Both the channel are of different intensities and and
> hence i am having a trouble getting a "yellow" merge image.

You should not want to get a yellow merge image. Our eyes are tuned to
discern shades of green in varying lighting conditions. So when we try to
quantify intensities, it is not a good idea to try that by eye.

If you think I am kidding, have a look at

http://fiji.sc/Colocalization_Analysis#Why_scatter_plots_instead_of_colour_merge_images.3F

The blue and the green are actually exactly the same color. If you think
that does not happen in your case because you use red/green channels,
download Fiji and start File>Open Samples>Spirals. You will see exactly
the same effect without any blue; instead the spirals look like green and
yellow, but still are of exactly the same color.

So you might think that by having a quick look you get an answer faster.
Sure, you get it quite fast, but the answer is also wrong.

Conclusion: never, ever, ever try to quantify intensities by eye.

There is another reason why you should not work with red/green merges: 9%
of the male population (which still make up the disproportionate majority
of scientists -- and reviewers!) cannot see the difference properly. It is
remarkable how many otherwise brilliant scientists forget that fact.

And lastly, there is an even more compelling reason why you should never,
ever rely on color merge images to prove your point: you cannot be sure
that the next person sees the same colors as you do. In fact, it has been
proven that color perception -- even outside of color blindness -- is very
subjective. And as scientists, we should try to stay away from
"subjective".

The good news is that it is actually very, very easy to do it the right
way, to quantify in an objective manner that other scientists have no
problem at reproducing: colocalization analysis.

With Fiji, we (mainly Daniel White) also tried to make colocalization
analysis not only easy but accessible: Download Fiji (for free, without
registration), follow the information and tutorials at

        http://fiji.sc/Colocalization_Analysis

and you're set.

Ciao,
Johannes