Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Loss of color diversity in LRGB photography when the filters do not overlap

  • Please log in to reply
292 replies to this topic

#26 James Peirce

James Peirce

    Viking 1

  • *****
  • Posts: 933
  • Joined: 21 Aug 2020
  • Loc: Salt Lake City, Utah, United States

Posted 12 March 2023 - 11:26 PM

Check it yourself. Do the experiment.

I know how you arrived at the result. You’re misinterpreting its meaning and representing the comparison in a way that bears no meaningful relationship. You’d have to capture and present the filtered CFA channels by the same criteria (blend channels for both if you will blend one) for the comparison to be helpful. Because you aren’t engaging with points made in discussion above, all I can do at this point is try to repeat or rephrase, which doesn’t seem productive. So instead I shall simply wish you well and move on.
  • imtl, steveincolo and hyiger like this

#27 imtl

imtl

    Cosmos

  • *****
  • Posts: 8,272
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 12 March 2023 - 11:59 PM

You seem to be misunderstanding something. Even so, you can see the effect in my pictures, and you can repeat these for yourself.  I am not saying that a spectrally pure yellow photon is lost, I am saying that it is mapped to green or red, so that no yellow will appear in the LRGB image. Take a picture of a spectrum with your own mono camera and nonoverlapping RGB filters. You won't get any yellow. You won't get any celeste or violet or orange either. It is physically impossible. You will just get plain red, plain green, and plain blue.

Yellow is a range of wavelengths. Indeed not a big range but it is not a single wavelength. It will be captured and it does show up with regular RGB filters (where there is a gap between G and R). Will OSC have a different color tone then RGB when capturing the yellow part of the spectrum? Sure. The filters are different. It does not make one match better then the other. And both of them do not match your color vision.



#28 arbit

arbit

    Apollo

  • -----
  • Posts: 1,396
  • Joined: 19 Feb 2012

Posted 13 March 2023 - 12:41 AM

I don't know where your yellow comes from, but if you read my comment carefully, maybe you can see whether my conditions apply to your situation.

 

 

I am not saying that a spectrally pure yellow photon is lost, I am saying that it is mapped to green or red, so that no yellow will appear in the LRGB image. Take a picture of a spectrum with your own mono camera and nonoverlapping RGB filters. You won't get any yellow. You won't get any celeste or violet or orange either. It is physically impossible. You will just get plain red, plain green, and plain blue.

Hi, loujost, I really don't understand this. You do get yellow (or orange for that matter), in any image you take with RGB. Just look at literally any RGB image of a galaxy core, or better still, Rho.

 

I don't know anything about colour science, but I do know the facts on seeing yellow in RGB images are pretty clear.

 

So if your hypothesis does hold, it would seem to be for a very specific situation and not generally for RGB imaging.


  • licho52 likes this

#29 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 07:52 AM

Hi, loujost, I really don't understand this. You do get yellow (or orange for that matter), in any image you take with RGB. Just look at literally any RGB image of a galaxy core, or better still, Rho.

 

I don't know anything about colour science, but I do know the facts on seeing yellow in RGB images are pretty clear.

 

So if your hypothesis does hold, it would seem to be for a very specific situation and not generally for RGB imaging.

You are right that "it would seem to be for a very specific situation."  I have repeatedly emphasized in my posts to this thread that the problem I am describing arises specifically for spectrally pure colors; colors like those in a rainbow, or an emission line, which fall into a narrow band of wavelengths. .Colors of stars and galaxies are caused by a different mechanism, blackbody radiation over a broad range of wavelengths, and those will be fine.Read what I wrote more carefully and I think you will understand. Almost no color science is needed.

 

 It's not a "hypothesis", it is just a simple fact. Anybody can check it empiricially.



#30 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 07:59 AM

Yellow is a range of wavelengths. Indeed not a big range but it is not a single wavelength. It will be captured and it does show up with regular RGB filters (where there is a gap between G and R). Will OSC have a different color tone then RGB when capturing the yellow part of the spectrum? Sure. The filters are different. It does not make one match better then the other. And both of them do not match your color vision.

I've shown by logic and by an example photo that spectrally pure yellow will be mapped to pure red or pure green in RGB photography on a monochrome camera, not a mixture of red and green. If you don't believe me, try it yourself with a  spectrally pure yellow (or a pure violet or a pure celeste or a pure orange). Or just think carefully about it.



#31 imtl

imtl

    Cosmos

  • *****
  • Posts: 8,272
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 13 March 2023 - 08:01 AM

I've shown by logic and by an example photo that spectrally pure yellow will be mapped to pure red or pure green in RGB photography on a monochrome camera, not a mixture of red and green. If you don't believe me, try it yourself with a  spectrally pure yellow (or a pure violet or a pure celeste or a pure orange). Or just think carefully about it.

Define pure yellow


  • hyiger likes this

#32 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 08:01 AM

I know how you arrived at the result. You’re misinterpreting its meaning and representing the comparison in a way that bears no meaningful relationship. You’d have to capture and present the filtered CFA channels by the same criteria (blend channels for both if you will blend one) for the comparison to be helpful. Because you aren’t engaging with points made in discussion above, all I can do at this point is try to repeat or rephrase, which doesn’t seem productive. So instead I shall simply wish you well and move on.

I AM blending the channels for both. When colors are spectrally pure and the filters do not overlap, for each pixel only one of the RGB channels contains any data.


Edited by loujost, 13 March 2023 - 08:08 AM.


#33 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 08:07 AM

Define pure yellow

A spectrally pure color is a narrowband color, like that of an emission line. A small range of wavelengths. In this context, "small" means being "falling completely within the wavelength range of one or the RGB filters" .  Because it only gets through one filter, it gets mapped directly to either red, green, or blue, not some combination.


Edited by loujost, 13 March 2023 - 08:07 AM.


#34 imtl

imtl

    Cosmos

  • *****
  • Posts: 8,272
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 13 March 2023 - 08:16 AM

A spectrally pure color is a narrowband color, like that of an emission line. A small range of wavelengths. In this context, "small" means being "falling completely within the wavelength range of one or the RGB filters" .  Because it only gets through one filter, it gets mapped directly to either red, green, or blue, not some combination.

That's not a definition of yellow. This is just arbitrarily deciding that "the range should small enough and skewed that it falls where I would like it to be, to prove my point".

 

I feel this discussion is not really going anywhere so I'm out. Maybe others would continue. Thanks.


  • steveincolo likes this

#35 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 08:51 AM

That's not a definition of yellow. This is just arbitrarily deciding that "the range should small enough and skewed that it falls where I would like it to be, to prove my point".

 

I feel this discussion is not really going anywhere so I'm out. Maybe others would continue. Thanks.

The problem isn't specifically with yellow, and applies to all narrowband colors imaged through filters with non-overlapping color ranges.. All will be mapped onto either R, G, or B. As you can see in the photo!!!

 

It's weird that some people don't get this, even with an actual photo that demonstrates the problem. One could legitimately argue that narrowband colors are not always important in AP, or things along that line, but I don't understand why people deny the basic principle, which is obviously true if given a moment's thought. The same thing happened in other CN discussions about this subject.

 

The monochrome RGB workflow, using RGB filters with non-overlapping spectral ranges, reduces the color diversity of an image, if more than one narrowband color falls in the spectral range of one of the R, G, or B filters. There is no problem for wide-spectrum black body radiators like stars.


Edited by loujost, 13 March 2023 - 10:32 AM.


#36 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 12:53 PM

Here are links to some other threads that reference this issue:

 

https://www.cloudyni...range-from-red/

 

https://www.cloudyni...numbers-please/

 

Both show the same pattern as in this current thread, people denying or misunderstanding the point even after repeated clear explanations from other posters, a  failure of good-faith engagement.

 

For example, in the first thread ("How do clean-cut LRGB filters tell spectral orange from red"), the OP and a few other participants have good comments (though of course he is trashed and misunderstood and criticized for many pages of incorrect comments---Comments #4, 5, 9 etc).

 

In the second thread, in Comment #14 a commenter persists in his erroneous interpretation. #15 clearly explains why he is wrong and explains that a rainbow would just contain blocks of red, green, and blue using LRGB imaging (and my picture proves this is correct). But the commenter just keeps right on denying in his Comment 16.

 

In that second thread, Mark Shelley (Comment #15) and Jon Rista (Comment #40) give good clear summaries of the issue I discuss here.

 

Anyway these threads may help others evaluate the results I have presented here (which should not be even slightly controversial).


Edited by loujost, 13 March 2023 - 12:58 PM.


#37 KBHornblower

KBHornblower

    Vanguard

  • -----
  • Posts: 2,302
  • Joined: 01 Jul 2020
  • Loc: Falls Church, VA (Washington DC suburb)

Posted 13 March 2023 - 01:07 PM

If I am not mistaken, a rainbow will not be as strongly impaired by non-overlapping filters as is the test spectrum in the opening post.  That is because the diameter of the Sun will smear the wavelengths around each gap into both adjacent bands.



#38 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 01:19 PM

If I am not mistaken, a rainbow will not be as strongly impaired by non-overlapping filters as is the test spectrum in the opening post.  That is because the diameter of the Sun will smear the wavelengths around each gap into both adjacent bands.

You may be right, I don't know enough to comment on that. However, from visual inspection, I think this effect is relatively small. Do you know the magnitude of this effect?



#39 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 01:34 PM

Yes you are right, several factors cause the rainbow to have slightly impure colors.

https://atoptics.co....ows/primcol.htm

My spectrum is cleaner.



#40 sharkmelley

sharkmelley

    Fly Me to the Moon

  • *****
  • Posts: 7,449
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 13 March 2023 - 04:03 PM

If you only capture RGB with typical astro filters that do not overlap, you will be unable to capture all the colors present in the scene, if some of those colors are spectrally pure.

I'm really pleased to see this example and I'm glad someone has actually performed this experiment!  

Up until now, it's just been a thought experiment that many people failed to comprehend.  Now there's no excuse!

 

Mark


  • jdupton, Jon Rista and colinrm like this

#41 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 04:06 PM

I'm really pleased to see this example and I'm glad someone has actually performed this experiment!  

Up until now, it's just been a thought experiment that many people failed to comprehend.  Now there's no excuse!

 

Mark

Many of them still don't believe it!!!!!!!


Edited by loujost, 13 March 2023 - 04:06 PM.

  • jdupton, sharkmelley, Jon Rista and 1 other like this

#42 jdupton

jdupton

    Fly Me to the Moon

  • *****
  • Posts: 5,551
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 13 March 2023 - 04:31 PM

Mark & loujost,

 

I'm really pleased to see this example and I'm glad someone has actually performed this experiment!  

Up until now, it's just been a thought experiment that many people failed to comprehend.  Now there's no excuse!

 

Mark

 

Many of them still don't believe it!!!!!!!

   I think a lot of information can get lost in the semantics of discussing the issue. This aspect has not come up specifically in this thread but in one of the many others it was discussed in terms of reproducing the colors using either Bayer or non-overlapping RGB filters.

 

   In my mind, it comes down to the question of how can one reproduce something that was never captured or measured in the first place? I see Bayer OSC as capturing all of the wavelengths in a manner that (sort of, usually, sometimes) resembles how the eye would respond. Information captured in such a system can then be reproduced in some manner for output. While a non-overlapping RGB filter set can create all of the missing colors in the gaps between filters, it is up to the user to synthesize those missing colors during processing.

 

   The two methods can both reproduce all of the colors of a rainbow but the Bayer method has at least captured some information about all the colors while the missing information from isolated RGB filters needs to be synthesized from what is being made available since some information was never captured.

 

   To me, this argument is more about the differences between reproducing colors and synthesizing colors from what information has been captured. I don't personally know of a way to reproduce something that was never captured / measured.

 

 

John


  • Jon Rista likes this

#43 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 04:40 PM

Mark & loujost,

 

 

   I think a lot of information can get lost in the semantics of discussing the issue. This aspect has not come up specifically in this thread but in one of the many others it was discussed in terms of reproducing the colors using either Bayer or non-overlapping RGB filters.

 

   In my mind, it comes down to the question of how can one reproduce something that was never captured or measured in the first place? I see Bayer OSC as capturing all of the wavelengths in a manner that (sort of, usually, sometimes) resembles how the eye would respond. Information captured in such a system can then be reproduced in some manner for output. While a non-overlapping RGB filter set can create all of the missing colors in the gaps between filters, it is up to the user to synthesize those missing colors during processing.

 

   The two methods can both reproduce all of the colors of a rainbow but the Bayer method has at least captured some information about all the colors while the missing information from isolated RGB filters needs to be synthesized from what is being made available since some information was never captured.

 

   To me, this argument is more about the differences between reproducing colors and synthesizing colors from what information has been captured. I don't personally know of a way to reproduce something that was never captured / measured.

 

 

John

John, I don't understand what you are saying.

"While a non-overlapping RGB filter set can create all of the missing colors in the gaps between filters, it is up to the user to synthesize those missing colors during processing."

 

I don't see how it can do this, but even if they could, I must emphasize that the effect we are discussing here is not caused by the gaps between filters. The effect would be the same even if there were no gaps between the filter spectra (as long as they do not actually overlap).



#44 jdupton

jdupton

    Fly Me to the Moon

  • *****
  • Posts: 5,551
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 13 March 2023 - 05:00 PM

loujost,

 

Judging from some previous threads on CN, some astrophotographers think that capturing RGB information is sufficient to reproduce all colors in the scene, because (almost) any color can be represented as an RGB triplet. After all, that's how monitors generate the colors we see in our photos.

   The case I was making is a subset of the same claim you cited in your opening Post here. Many think that since RGB color layers can be combined to produce all possible display colors, RGB is capturing all the information needed for any scene. I read once here on CN that the claim was extended to narrowband lines in the spectrum -- a 5 nM filter in the Red, Green, and Blue portions of the spectrum could be used to create all the colors needed for image display. Since each line then comprises the whole color channel, a rainbow might look like the second image in your original post even when just a tiny slice of the spectrum was being sampled. 

 

 

John



#45 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 05:09 PM

loujost,

 

   The case I was making is a subset of the same claim you cited in your opening Post here. Many think that since RGB color layers can be combined to produce all possible display colors, RGB is capturing all the information needed for any scene. I read once here on CN that the claim was extended to narrowband lines in the spectrum -- a 5 nM filter in the Red, Green, and Blue portions of the spectrum could be used to create all the colors needed for image display. Since each line then comprises the whole color channel, a rainbow might look like the second image in your original post even when just a tiny slice of the spectrum was being sampled. 

 

 

John

Yes, thanks for the clarification.



#46 sharkmelley

sharkmelley

    Fly Me to the Moon

  • *****
  • Posts: 7,449
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 13 March 2023 - 05:18 PM

I think a lot of information can get lost in the semantics of discussing the issue. 

Agreed.  The discussion about "yellow" is a perfect example of this, especially since monochromatic yellow (580nm) is deliberately in the gap between red and green filters in some filter sets (e.g. Astronomik/Astrodon) so it's not recorded at all!

As I've pointed out on previous occasions, this caused problems for people trying to record the colour of the Sodium tail of Comet Neowise.


  • Jon Rista likes this

#47 KBHornblower

KBHornblower

    Vanguard

  • -----
  • Posts: 2,302
  • Joined: 01 Jul 2020
  • Loc: Falls Church, VA (Washington DC suburb)

Posted 13 March 2023 - 06:54 PM

Out of curiosity I tested the camera in my iPhone 13 on a broad spectrum created by a transmission grating.  The light source is the LED headlight on my electric lawn mower, viewed from an angle that makes it look like a narrow slit.

 

Spectrum iPhone reduced.jpg

 

This camera clearly has little or no overlap, and in addition it has a sharp cutoff in the blue band.  Visually, the blue part of the spectrum is just as wide as the green, and trails off into the violet end looking distinctly purple.  This camera works beautifully on faces and ordinary scenes which have broad spectra, but it would be unsatisfactory for rendering emission-line sources.

 

While I am at it, here is a rainbow.

 

Rainbow over Jaguarland reduced.jpg

 

This was copied from a friend's files, so I do not know anything about the camera.



#48 loujost

loujost

    Mariner 2

  • -----
  • topic starter
  • Posts: 243
  • Joined: 02 Oct 2017

Posted 13 March 2023 - 07:14 PM

Out of curiosity I tested the camera in my iPhone 13 on a broad spectrum created by a transmission grating.  The light source is the LED headlight on my electric lawn mower, viewed from an angle that makes it look like a narrow slit.

 

attachicon.gifSpectrum iPhone reduced.jpg

 

This camera clearly has little or no overlap, and in addition it has a sharp cutoff in the blue band.  Visually, the blue part of the spectrum is just as wide as the green, and trails off into the violet end looking distinctly purple.  This camera works beautifully on faces and ordinary scenes which have broad spectra, but it would be unsatisfactory for rendering emission-line sources.

 

While I am at it, here is a rainbow.

 

attachicon.gifRainbow over Jaguarland reduced.jpg

 

This was copied from a friend's files, so I do not know anything about the camera.

Thanks for the interesting information. So much yellow in that rainbow compared to the spectrum.


Edited by loujost, 13 March 2023 - 07:15 PM.


#49 KBHornblower

KBHornblower

    Vanguard

  • -----
  • Posts: 2,302
  • Joined: 01 Jul 2020
  • Loc: Falls Church, VA (Washington DC suburb)

Posted 13 March 2023 - 07:53 PM

Addendum:  Here is the same test spectrum as before, but with my 17-year-old Fuji FinePix compact camera.

 

Spectrum Fuji reduced.jpg

 

Unlike the phone camera, this one actually displays the far violet as distinctly purple, which can be displayed only by using a mixture of blue and red in the monitor.  My educated guess is that the red filter in the camera is also transmitting the violet light.  If I am not mistaken, the L cones in our eyes do the same thing.  The dimness of the blue is a characteristic of the white LED.  Visually, that part of the spectrum is distinctly brighter with an incandescent filament.

 

My guess is that Apple chose to cut off more of the violet end of the spectrum to be sure of stopping ultraviolet that could fog the image.  Apparently that part of the spectrum contributes very little to ordinary scenes on which the phone camera does just fine.



#50 KBHornblower

KBHornblower

    Vanguard

  • -----
  • Posts: 2,302
  • Joined: 01 Jul 2020
  • Loc: Falls Church, VA (Washington DC suburb)

Posted 13 March 2023 - 08:04 PM

Thanks for the interesting information. So much yellow in that rainbow compared to the spectrum.

Once again, that is because of the finite angular size of the Sun.  The overall dispersion in what we can see of the rainbow is only about 2 degrees, while light of any given color is projected over a span of 1/2 degree, giving a lot of blending.  I see the same effect when looking at a frosted light bulb through my grating.




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics