Transmitted resolution with Bayer filter color cameras

[G. Esteban Fernandez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

I drill it into my users to not take fluorescence pictures with color
(Bayer filter) cameras because of the loss in resolution.  Today
someone asked me if the same is true for transmitted images and I was
at a loss.  I can reason that for a somewhat pure R, G, or B color,
like blue staining for B-gal, there might be a loss in res. in a Bayer
image compared to monochrome but what about for brownish or black
staining?  And is there really a loss even for something that is
"blue" but also has R and G components, albeit at much less contrast?

Thanks,
Esteban

[mmodel]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Esteban,

This is not exactly an answer to your question, but I think that theoretical resolution is seldom achieved in transmitted images of sections because of their thickness and out-of-focus blur. It seems to me that more important than achieving the best possible resolution it to be able to clearly see just what you want to see -using contrast, staining or whatever it takes... With that said, I sometimes take images of single-stained sections using a monochrome camera through a filter that would maximize their contrast - if the monochrome camera is of better quality and gives me less noise.

Mike

________________________________________
From: Confocal Microscopy List [[hidden email]] On Behalf Of G. Esteban Fernandez [[hidden email]]
Sent: Tuesday, August 30, 2011 5:58 PM
To: [hidden email]
Subject: Transmitted resolution with Bayer filter color cameras

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

I drill it into my users to not take fluorescence pictures with color
(Bayer filter) cameras because of the loss in resolution.  Today
someone asked me if the same is true for transmitted images and I was
at a loss.  I can reason that for a somewhat pure R, G, or B color,
like blue staining for B-gal, there might be a loss in res. in a Bayer
image compared to monochrome but what about for brownish or black
staining?  And is there really a loss even for something that is
"blue" but also has R and G components, albeit at much less contrast?

Thanks,
Esteban

[Guy Cox-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

The key question in the end is whether you meet the Nyquist criterion
bearing in mind that with a Bayer mosaic your resel is larger than your
pixel.  Many Bayer mosaic cameras are multi-megapixel and easily meet
this criterion.  If this is so then they are fine for both fluorescence
and transmitted light - I've certainly used Bayer cameras for both
techniques.

                                Guy

Optical Imaging Techniques in Cell Biology
by Guy Cox    CRC Press / Taylor & Francis
     http://www.guycox.com/optical.htm
______________________________________________
Associate Professor Guy Cox, MA, DPhil(Oxon)
Australian Centre for Microscopy & Microanalysis,
Madsen Building F09, University of Sydney, NSW 2006

Phone [phone]     Fax [phone]
             Mobile [phone]
______________________________________________
      http://www.guycox.net
 


-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]]
On Behalf Of G. Esteban Fernandez
Sent: Wednesday, 31 August 2011 7:59 AM
To: [hidden email]
Subject: Transmitted resolution with Bayer filter color cameras

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

I drill it into my users to not take fluorescence pictures with color
(Bayer filter) cameras because of the loss in resolution.  Today
someone asked me if the same is true for transmitted images and I was
at a loss.  I can reason that for a somewhat pure R, G, or B color,
like blue staining for B-gal, there might be a loss in res. in a Bayer
image compared to monochrome but what about for brownish or black
staining?  And is there really a loss even for something that is
"blue" but also has R and G components, albeit at much less contrast?

Thanks,
Esteban

[G. Esteban Fernandez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I know I'm not meeting Nyquist even on a single pixel level (without
the Bayer filter).  It is an AxioCam MRc of 1.4 MP.  The same 'scope
also has the exact same camera but without the Bayer filter, AxioCam
MRm.  Users easily switch between the two and I want know how much
res, if any, they're losing by taking transmitted images in color.  Of
course there is a huge contrast benefit from color for most specimens
but not for many that are dark brown staining or one color.

So I guess my question really is then, what's the resel size in a
Bayer camera for white light?

Thanks,
Esteban


On Tue, Aug 30, 2011 at 8:25 PM, Guy Cox <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> The key question in the end is whether you meet the Nyquist criterion
> bearing in mind that with a Bayer mosaic your resel is larger than your
> pixel.  Many Bayer mosaic cameras are multi-megapixel and easily meet
> this criterion.  If this is so then they are fine for both fluorescence
> and transmitted light - I've certainly used Bayer cameras for both
> techniques.
>
>                                Guy
>
> Optical Imaging Techniques in Cell Biology
> by Guy Cox    CRC Press / Taylor & Francis
>     http://www.guycox.com/optical.htm
> ______________________________________________
> Associate Professor Guy Cox, MA, DPhil(Oxon)
> Australian Centre for Microscopy & Microanalysis,
> Madsen Building F09, University of Sydney, NSW 2006
>
> Phone [phone]     Fax [phone]
>             Mobile [phone]
> ______________________________________________
>      http://www.guycox.net
>
>
>
> -----Original Message-----
> From: Confocal Microscopy List [mailto:[hidden email]]
> On Behalf Of G. Esteban Fernandez
> Sent: Wednesday, 31 August 2011 7:59 AM
> To: [hidden email]
> Subject: Transmitted resolution with Bayer filter color cameras
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi all,
>
> I drill it into my users to not take fluorescence pictures with color
> (Bayer filter) cameras because of the loss in resolution.  Today
> someone asked me if the same is true for transmitted images and I was
> at a loss.  I can reason that for a somewhat pure R, G, or B color,
> like blue staining for B-gal, there might be a loss in res. in a Bayer
> image compared to monochrome but what about for brownish or black
> staining?  And is there really a loss even for something that is
> "blue" but also has R and G components, albeit at much less contrast?
>
> Thanks,
> Esteban
>

[Phillips, Thomas E.]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I think taking a B&W transmitted light photo under optimal conditions would beat out a color Bayer image since using a blue filter in the excitation path would increase your resolution. If you had a poorly corrected achromat, a green filter would help. In regards to loss of contrast with B&W, back in the dark ages before color micrographs were common in scientific publications, it was common to use a complementary color excitation filter to increase the contrast of specific stains in B&W photos. I have a whole series of Wratten filters just for that purpose collecting dust in some cabinet.


Thomas E. Phillips, Ph.D
Professor of Biological Sciences
Director, Molecular Cytology Core
2 Tucker Hall
University of Missouri
Columbia, MO 65211-7400
573-882-4712 (office)
573-882-0123 (fax)
[hidden email]

http://www.biology.missouri.edu/faculty/phillips.html
http://www.biotech.missouri.edu/mcc/


-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of G. Esteban Fernandez
Sent: Tuesday, August 30, 2011 4:59 PM
To: [hidden email]
Subject: Transmitted resolution with Bayer filter color cameras

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

I drill it into my users to not take fluorescence pictures with color
(Bayer filter) cameras because of the loss in resolution.  Today
someone asked me if the same is true for transmitted images and I was
at a loss.  I can reason that for a somewhat pure R, G, or B color,
like blue staining for B-gal, there might be a loss in res. in a Bayer
image compared to monochrome but what about for brownish or black
staining?  And is there really a loss even for something that is
"blue" but also has R and G components, albeit at much less contrast?

Thanks,
Esteban

[Julio Vazquez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I would guess that if you are using white light, it shouldn't matter a great deal whether you use a color or monochrome camera, since all pixels will be recording information, so the only thing you may loose is a bit of quantum efficiency, which in transmitted light isn't a big deal. However, if your sample happens to have well-defined colors, such as red, then the information will be recorded primarily on a subset of pixels, with the rest being interpolated, and the comments below would apply.

If you are using the color camera for imaging fluorescence (or transmitted light images with narrow range of colors), then the issue is that intensity information will be recorded primarily on a subset of pixels (such as red pixels for a red emitting dye), with much lower intensity levels recorded by the other pixels, the amount of which will depend on exactly which wavelengths your dye is emitting at, and how much of that is transmitted and being collected by the pixels under the Red, Green, and Blue filters. Then, for each pixel, the values for the other two colors will be interpolated. In a typical Bayer mask, for each set of 2x2 pixels, there will be two Green, 1 Red and 1 Blue, so there will be more interpolation for Red and Blue, than for Green. It also seems to me that if the color you are imaging is close to a primary, the amount of interpolated information will be greater (because the values will need to be computed for the remaining two colors). On the other hand, if the color you are imaging is not a primary, information will be more evenly distributed (for example, a yellow structure will be recorded to comparable degree on the Green and Red pixels, and interpolation will be primarily for the Blue pixel (i.e. 1 out of four pixels). In fact, if you look at the spec sheet, the bands of detection are pretty broad. In particular, the response for the  "Green" pixels is pretty broad, which means that for most of the commonly used spectrum, you are typically collecting information on three out of four of your RGB pixels.

I wouldn't know how to do the type of math you are asking, but my guess again is that if you are sampling correctly, the difference in resolution should be quite modest, if noticeable at all. If you are undersampling, then I would assume that if you are imaging a diffraction-limited structure, and that your sampling is such that the image of that point-like structure falls on the wrong pixel (e.g. red structure fall on "Blue" pixel, then you will hardly get any signal or information, and that structure will pretty much be invisible.

My take on this is that if there is a mathematical answer to your question, it is bound to be not trivial.... my intuition tells me that the resolution you achieve on a bayer mask camera will depend on the color you are imaging and your sampling, and will affect primarily very small structures at or below the pixel size. However, since you have both cameras, it should be easy to take two images of a suitable sample, and determine for yourself whether you can see any difference. I have occasionally used color cameras to take fluorescence images (and know many labs who do that because at some point they had to choose a single camera for their scope and went with color), and I was actually surprised that they were quite good. If I am imaging a red dye, I just take a color image, separate the channels, and keep the red. You could then compare this to the same image acquired on the monochrome camera.

Finally, I don't think this applies to the Axiocam MRc, but I believe the Axiocam HRc can shift the Bayer mask around and collect all color information at all pixels, in which case, except for the light loss of QE due to the filters, it should be quite similar to a monochrome camera.


--
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA


http://www.fhcrc.org/science/shared_resources/imaging/

==

On Aug 31, 2011, at 10:18 AM, G. Esteban Fernandez wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> I know I'm not meeting Nyquist even on a single pixel level (without
> the Bayer filter).  It is an AxioCam MRc of 1.4 MP.  The same 'scope
> also has the exact same camera but without the Bayer filter, AxioCam
> MRm.  Users easily switch between the two and I want know how much
> res, if any, they're losing by taking transmitted images in color.  Of
> course there is a huge contrast benefit from color for most specimens
> but not for many that are dark brown staining or one color.
>
> So I guess my question really is then, what's the resel size in a
> Bayer camera for white light?
>
> Thanks,
> Esteban

[G. Esteban Fernandez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

You're right, ultimately for me it comes down to the users, so no
matter what the theoretical res. is I should just take pics with both
cameras to show users the differece.

Thanks everybody.

-Esteban


On Wed, Aug 31, 2011 at 11:14 AM, Julio Vazquez <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> I would guess that if you are using white light, it shouldn't matter a great deal whether you use a color or monochrome camera, since all pixels will be recording information, so the only thing you may loose is a bit of quantum efficiency, which in transmitted light isn't a big deal. However, if your sample happens to have well-defined colors, such as red, then the information will be recorded primarily on a subset of pixels, with the rest being interpolated, and the comments below would apply.
>
> If you are using the color camera for imaging fluorescence (or transmitted light images with narrow range of colors), then the issue is that intensity information will be recorded primarily on a subset of pixels (such as red pixels for a red emitting dye), with much lower intensity levels recorded by the other pixels, the amount of which will depend on exactly which wavelengths your dye is emitting at, and how much of that is transmitted and being collected by the pixels under the Red, Green, and Blue filters. Then, for each pixel, the values for the other two colors will be interpolated. In a typical Bayer mask, for each set of 2x2 pixels, there will be two Green, 1 Red and 1 Blue, so there will be more interpolation for Red and Blue, than for Green. It also seems to me that if the color you are imaging is close to a primary, the amount of interpolated information will be greater (because the values will need to be computed for the remaining two colors). On the other hand, if the color you are imaging is not a primary, information will be more evenly distributed (for example, a yellow structure will be recorded to comparable degree on the Green and Red pixels, and interpolation will be primarily for the Blue pixel (i.e. 1 out of four pixels). In fact, if you look at the spec sheet, the bands of detection are pretty broad. In particular, the response for the  "Green" pixels is pretty broad, which means that for most of the commonly used spectrum, you are typically collecting information on three out of four of your RGB pixels.
>
> I wouldn't know how to do the type of math you are asking, but my guess again is that if you are sampling correctly, the difference in resolution should be quite modest, if noticeable at all. If you are undersampling, then I would assume that if you are imaging a diffraction-limited structure, and that your sampling is such that the image of that point-like structure falls on the wrong pixel (e.g. red structure fall on "Blue" pixel, then you will hardly get any signal or information, and that structure will pretty much be invisible.
>
> My take on this is that if there is a mathematical answer to your question, it is bound to be not trivial.... my intuition tells me that the resolution you achieve on a bayer mask camera will depend on the color you are imaging and your sampling, and will affect primarily very small structures at or below the pixel size. However, since you have both cameras, it should be easy to take two images of a suitable sample, and determine for yourself whether you can see any difference. I have occasionally used color cameras to take fluorescence images (and know many labs who do that because at some point they had to choose a single camera for their scope and went with color), and I was actually surprised that they were quite good. If I am imaging a red dye, I just take a color image, separate the channels, and keep the red. You could then compare this to the same image acquired on the monochrome camera.
>
> Finally, I don't think this applies to the Axiocam MRc, but I believe the Axiocam HRc can shift the Bayer mask around and collect all color information at all pixels, in which case, except for the light loss of QE due to the filters, it should be quite similar to a monochrome camera.
>
>
> --
> Julio Vazquez
> Fred Hutchinson Cancer Research Center
> Seattle, WA
>
>
> http://www.fhcrc.org/science/shared_resources/imaging/
>
> ==
>
> On Aug 31, 2011, at 10:18 AM, G. Esteban Fernandez wrote:
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> I know I'm not meeting Nyquist even on a single pixel level (without
>> the Bayer filter).  It is an AxioCam MRc of 1.4 MP.  The same 'scope
>> also has the exact same camera but without the Bayer filter, AxioCam
>> MRm.  Users easily switch between the two and I want know how much
>> res, if any, they're losing by taking transmitted images in color.  Of
>> course there is a huge contrast benefit from color for most specimens
>> but not for many that are dark brown staining or one color.
>>
>> So I guess my question really is then, what's the resel size in a
>> Bayer camera for white light?
>>
>> Thanks,
>> Esteban
>