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

#101 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 26,034
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 15 June 2023 - 09:58 AM

Jon, that's pretty close, but most references also show a bump in red sensitivity in deep blue. This is what produces violet

 

So, I used to think the same...but, I believe those plots with the bump are actually CIE 1931 color matching functions. In fact, there are two...the XYZ color matching functions have the bump of red in the blues, the CIE RGB color matching functions actually have a drop. 

 

The plot I shared above, is the actual sensitivities of the S (short, blue), M (medium, green) and L (long, red) wavelength sensitivities of human cone cells. So the plot shared above should actually represent natural human vision sensitivities. I have found some other plots of the same, however they, too, all have the same peak heights. The labels state they were "normalized" plots, so I am not sure exactly what kind of normalization was done.


Edited by Jon Rista, 15 June 2023 - 10:09 AM.


#102 loujost

loujost

    Mariner 2

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

Posted 15 June 2023 - 12:45 PM

So, I used to think the same...but, I believe those plots with the bump are actually CIE 1931 color matching functions. In fact, there are two...the XYZ color matching functions have the bump of red in the blues, the CIE RGB color matching functions actually have a drop. 

 

The plot I shared above, is the actual sensitivities of the S (short, blue), M (medium, green) and L (long, red) wavelength sensitivities of human cone cells. So the plot shared above should actually represent natural human vision sensitivities. I have found some other plots of the same, however they, too, all have the same peak heights. The labels state they were "normalized" plots, so I am not sure exactly what kind of normalization was done.

But if that were true , how is violet generated?



#103 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 26,034
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 16 June 2023 - 08:48 PM

But if that were true , how is violet generated?

There is still overlap of red and blue...there just isn't a secondary peak of sensitivity of the red cones, in the blues. 



#104 loujost

loujost

    Mariner 2

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

Posted 16 June 2023 - 09:15 PM

There is still overlap of red and blue...there just isn't a secondary peak of sensitivity of the red cones, in the blues. 

But in your graph the red response decreases (both in relative and absolute values) as the stimulus decreases in wavelength. Our violet response does the opposite. There has to be a bump in the red response in order to produce the observed effect..
 


Edited by loujost, 16 June 2023 - 09:16 PM.


#105 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 16 June 2023 - 11:08 PM

But in your graph the red response decreases (both in relative and absolute values) as the stimulus decreases in wavelength. Our violet response does the opposite. There has to be a bump in the red response in order to produce the observed effect..
 

I think it's better to say we perceive the non-spectral color purple as violet because there is a blip of red perceptual response that overlaps with blue in the violet range.

 

I see lots of response curves on the web that show a blip of response for red in the deep blue - but many others don't show it.  Many of the plots look cartoonish and hand-drawn and have no cited source.

 

I think the difference is that human cone spectral response for LMS cones does *not* show the blip for the L cones, but nonetheless the color-matching functions used for CIE XYZ *do* show that small peak for the X coordinate.  And that curve is based on perceptual tests with many humans with normal vision.

So I guess the biophysical cone response doesn't have the peak, but the brain puts one there - or something.  It's as if the gain is turned up on red down there so that violet colors don't simply register as deep blue.  As a consequence, a combination of red and normal blue is perceived as violet.

 

There is a good summary of all this in Handbook of Optics vol. 1.

 

Frank


  • Jon Rista likes this

#106 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 26,034
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 17 June 2023 - 12:01 PM

But in your graph the red response decreases (both in relative and absolute values) as the stimulus decreases in wavelength. Our violet response does the opposite. There has to be a bump in the red response in order to produce the observed effect..
 

All I know, is that one set of curves is analytical, based on measured responses. The others are computed. It is the XYZ color matching matrix that has the bump of red in the blues. I don't know why, but it is a color  transformation thing into/out of that color model. The RGB color matching matrix actually has a drop in red, right around the blues, rather than a bump. The curves I posted, as far as I understand, are based on the analysis of actual responses of human cone cells to light. 



#107 badgie

badgie

    Viking 1

  • *****
  • Posts: 643
  • Joined: 13 Jan 2021
  • Loc: SF Area, California

Posted 17 June 2023 - 06:19 PM

I'm curious here about optimal filters for color accuracy. If we stated as a goal the ability to accurately identify any single wavelength then we'd be looking at something the one filter sloping up and one down with full overlap. Without filter overlap other isn't enough info to distinguish different colors. For non trivial colors (non single wavelengths) this is a more complex question. Does anyone know of optimal (information theoretical) filters for this scenario?

#108 loujost

loujost

    Mariner 2

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

Posted 17 June 2023 - 07:55 PM

I'm sorry but I just cannot understand how those curves could be correct.

 

"I think it's better to say we perceive the non-spectral color purple as violet because there is a blip of red perceptual response that overlaps with blue in the violet range."

 

But Violet is a spectrally pure color; look at any rainbow or spectroscope. or look at a 405nm laser.  You mention a blip of red response as if it were a sufficient explanation for perception of violet, but as I keep mentioning, that theory is disproven by the fact that we perceive violet for the SHORTER wavelengths, not the wavelengths where, in the "no-bump" graphs, the relative amount of red response is allegedly greatest, at the longer wavelengths of blue light.

 

Jon, you say "There is still overlap of red and blue...there just isn't a secondary peak of sensitivity of the red cones, in the blues." Alright, but a complete explanation has to explain why we see violet at the shorter end of the blue wavelengths, and not at the longer end of the blue wavelengths. I suppose there doesn't need to be a second pronounced peak (a local maximum) in red response, but there must be a bump of some kind, so that there is relatively more Red than Blue stimulus at the shorter blue wavelengths, versus the relative amount of Red and Blue stimulus at longer blue wavelengths. 


Edited by loujost, 17 June 2023 - 07:56 PM.


#109 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 17 June 2023 - 08:37 PM

I'm sorry but I just cannot understand how those curves could be correct.

 

"I think it's better to say we perceive the non-spectral color purple as violet because there is a blip of red perceptual response that overlaps with blue in the violet range."

 

But Violet is a spectrally pure color; look at any rainbow or spectroscope. or look at a 405nm laser.  You mention a blip of red response as if it were a sufficient explanation for perception of violet, but as I keep mentioning, that theory is disproven by the fact that we perceive violet for the SHORTER wavelengths, not the wavelengths where, in the "no-bump" graphs, the relative amount of red response is allegedly greatest, at the longer wavelengths of blue light.

 

Jon, you say "There is still overlap of red and blue...there just isn't a secondary peak of sensitivity of the red cones, in the blues." Alright, but a complete explanation has to explain why we see violet at the shorter end of the blue wavelengths, and not at the longer end of the blue wavelengths. I suppose there doesn't need to be a second pronounced peak (a local maximum) in red response, but there must be a bump of some kind, so that there is relatively more Red than Blue stimulus at the shorter blue wavelengths, versus the relative amount of Red and Blue stimulus at longer blue wavelengths. 

Violet is a spectral color but purple is not.  We see violet as violet - and we see a combination of red and blue as violet.  That is how the CIE response curves were created.  You are given R, G, B lights with dials on them that combine to create a single color, and you set the dials to match a given wavelength shown separately.  It is the setting of the dials that results in the curves shown with the red bump on the left - and the odd thing is that the cones do *not* show such a blip.  But cones combined with the human perceptual response say that the violet color is triggered when the cones receive a mixture of red and blue.  (It is also triggered when we receive violet light).  That's just the way it is - but it's fortunate because it means using only R, G, B pixels we can generate a response that appears violet rather than blue - if we toss in a little red. 

 

So how do we know that a sensor pixel is receiving violet rather than pure blue light?  Well - we give red a bit of response down there and it all happens automatically.  The signal will be recorded as a mixture of blue and red - and it will then be displayed and perceived that way - even though there are no violet sensor pixels and no violet led pixels.  It's all done with R, G, B.

 

Frank


Edited by freestar8n, 17 June 2023 - 08:40 PM.


#110 loujost

loujost

    Mariner 2

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

Posted 17 June 2023 - 08:47 PM

Violet is a spectral color but purple is not.  We see violet as violet - and we see a combination of red and blue as violet.  That is how the CIE response curves were created.  You are given R, G, B lights with dials on them that combine to create a single color, and you set the dials to match a given wavelength shown separately.  It is the setting of the dials that results in the curves shown with the red bump on the left - and the odd thing is that the cones do *not* show such a blip.  But cones combined with the human perceptual response say that the violet color is triggered when the cones receive a mixture of red and blue.  (It is also triggered when we receive violet light).  That's just the way it is - but it's fortunate because it means using only R, G, B pixels we can generate a response that appears violet rather than blue - if we toss in a little red. 

 

So how do we know that a sensor pixel is receiving violet rather than pure blue light?  Well - we give red a bit of response down there and it all happens automatically.  The signal will be recorded as a mixture of blue and red - and it will then be displayed and perceived that way - even though there are no violet sensor pixels and no violet led pixels.  It's all done with R, G, B.

 

Frank

But none of that addresses my point at all. We perceive violet at the short end of the blue spectrum but not at the long end. Your explanation does not predict that.



#111 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 17 June 2023 - 09:24 PM

But none of that addresses my point at all. We perceive violet at the short end of the blue spectrum but not at the long end. Your explanation does not predict that.

?  The relative quantities matter of course.  Strong blue and weak red is violet.  Strong red and weak blue is red.  Strong blue and no red is blue.

 

Your brain can do whatever it wants with the signals it receives - and apparently this is what it does.  As reflected by the CIE curves, and the fact that Bayer filters and LED displays do so well across the spectrum.

 

With completely separate R and B bandpasses in mono filters I don't know how you would distinguish violet from blue.  But there isn't a lot of violet in the night sky.  There is purplish, though - like the Ha/HB combo - and that works ok.  I haven't seen a mono camera and mono filters try to image a rainbow.  I don't know how it could render the violet part - it should just stop at blue.

 

Frank



#112 loujost

loujost

    Mariner 2

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

Posted 17 June 2023 - 10:20 PM

?  The relative quantities matter of course.  Strong blue and weak red is violet.  Strong red and weak blue is red.  Strong blue and no red is blue.

 

Your brain can do whatever it wants with the signals it receives - and apparently this is what it does.  As reflected by the CIE curves, and the fact that Bayer filters and LED displays do so well across the spectrum.

 

With completely separate R and B bandpasses in mono filters I don't know how you would distinguish violet from blue.  But there isn't a lot of violet in the night sky.  There is purplish, though - like the Ha/HB combo - and that works ok.  I haven't seen a mono camera and mono filters try to image a rainbow.  I don't know how it could render the violet part - it should just stop at blue.

 

Frank

That last part is right, as you can see in my photo at the start of this thread. Non-overlapping RGB filters cannot record any violet. That's why I gave this thread its title.

But there is still something missing in the explanation for visual violet.



#113 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 17 June 2023 - 11:04 PM

That last part is right, as you can see in my photo at the start of this thread. Non-overlapping RGB filters cannot record any violet. That's why I gave this thread its title.

But there is still something missing in the explanation for visual violet.

Yes I forgot the first part of this thread.  But as for violet I don't see anything missing.  We register it as blue with a bit of red - and that translates to violet.

 

Rainbows and spectra are special in that each point maps to a single wavelength.  As a result, with non-overlapping filters you would only record each color with a single filter.  But that doesn't mean you can't faithfully capture yellow or orange - because such colors in a scene will likely be due to broad spectrum sources - rather than a single emission line.  So that color will likely register on 2 or 3 filters to produce a triad.  And with good color calibration of the scene - it should be rendered accurately.

 

It doesn't require overlapping filters for that to work.  But it may be an issue with narrow emission lines.  If you want every wavelength to record and display properly by itself - yes you would need overlap and similarity of the response/display to human vision.  And as part of that you would need to record violet as blue combined with a bit of red.  I think Bayer filters do pretty well with all that.  You can look at images of ionized gases on the web taken with conventional cameras - and they all look pretty good to me.  So do images of a Macbeth chart.

 

Frank



#114 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 07:21 AM

Frank, yes, I agree with most of that. That's what this trhead is about- loss of color diversity due to mapping of all emission lines onto the same shades of R, G, and B.

 

But I can't seem to convince you of the problem with your explanation of violet in terms of bumpless sensitiviity curves. Look at your explanation carefully. It is as if you aren't really looking at the graphs.

 

"We register it as blue with a bit of red - and that translates to violet."

 

If that were the whole story, and the curves in Post #99 were correct, ask yourself where is the most red response in the blue region. It's in the region of celeste, not violet.

 

In the region where we actually see violet, the red curve shows a negligible response, approaching zero.

 

Maybe the green response cancels out the red response in the  celeste region; that does appear in some published curves. But it still seems like you need a bump in the red curve (maybe not a local maximum, but some kind of step or swelling) to make enough red response in the region where we really do see violet.


  • Dan_I likes this

#115 badgie

badgie

    Viking 1

  • *****
  • Posts: 643
  • Joined: 13 Jan 2021
  • Loc: SF Area, California

Posted 18 June 2023 - 12:28 PM

It's not completely clear to me what the state of the literature is on this.  Looking around, many cite Bowmaker JK, Dartnall HJ. Visual pigments of rods and cones in a human retina. J Physiol. 1980 Jan;298:501-11. doi: 10.1113/jphysiol.1980.sp013097 which does appear to show a slight increase in response of red cones from 450 -> 400 based on transmission spectroscopy. 

 

It looks like the more modern references don't show this effect.  This website (http://cvrl.ucl.ac.uk/cones.htm) shows many different response curves, none of which show an increase in red response at the bluest end of the spectrum.  But these plots end at about 390 so maybe you need to go deeper into the violet?



#116 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 01:04 PM

Surely someone must have measured directly the response curve of an L (red) cone exposed to spectrally pure light at different wavelengths...



#117 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 01:10 PM

https://www.unm.edu/...ne_response.htm

 

This is supposedly the action spectrum of each type of cone (and rods), so I would think this is actually the measured physical response of the cones, with no color matching involved. Note the red bump right where it really needs to be in order to explain our perception of violet.



#118 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 01:21 PM

https://webvision.me...ourasFig 14.jpg

 

Again here we see the physical absorption spectrum for each type of cone; no color matching involved. It again shows a red bump right where it MUST be in order to really explain our perception of violet.


Edited by loujost, 18 June 2023 - 01:25 PM.


#119 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 26,034
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 18 June 2023 - 01:38 PM

I'm sorry but I just cannot understand how those curves could be correct.

 

"I think it's better to say we perceive the non-spectral color purple as violet because there is a blip of red perceptual response that overlaps with blue in the violet range."

 

But Violet is a spectrally pure color; look at any rainbow or spectroscope. or look at a 405nm laser.  You mention a blip of red response as if it were a sufficient explanation for perception of violet, but as I keep mentioning, that theory is disproven by the fact that we perceive violet for the SHORTER wavelengths, not the wavelengths where, in the "no-bump" graphs, the relative amount of red response is allegedly greatest, at the longer wavelengths of blue light.

 

Jon, you say "There is still overlap of red and blue...there just isn't a secondary peak of sensitivity of the red cones, in the blues." Alright, but a complete explanation has to explain why we see violet at the shorter end of the blue wavelengths, and not at the longer end of the blue wavelengths. I suppose there doesn't need to be a second pronounced peak (a local maximum) in red response, but there must be a bump of some kind, so that there is relatively more Red than Blue stimulus at the shorter blue wavelengths, versus the relative amount of Red and Blue stimulus at longer blue wavelengths. 

As you stated, violet is a spectrally pure color. Look at the response of the blue cones...it extends into the Near UV range. As such, would blue cones themselves, not be responsible for our sensitivity to violet light, a spectrally pure color?? Why is a red response required here?


Edited by Jon Rista, 18 June 2023 - 01:38 PM.


#120 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 26,034
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 18 June 2023 - 01:43 PM

https://www.unm.edu/...ne_response.htm

 

This is supposedly the action spectrum of each type of cone (and rods), so I would think this is actually the measured physical response of the cones, with no color matching involved. Note the red bump right where it really needs to be in order to explain our perception of violet.

 

https://webvision.me...ourasFig 14.jpg

 

Again here we see the physical absorption spectrum for each type of cone; no color matching involved. It again shows a red bump right where it MUST be in order to really explain our perception of violet.

Ok! So this explains why the other plots I found were called "Relative"...because they were not absolute plots of measured response. I don't know why the relative plots are so easy to find...

 

Still...I query, looking at the S cone response, their peak response is to violet. Some studies call the red cones "magenta" cones, due to their response, which I believe is why the red bump in the blues is REALLY necessary. Why couldn't a blue cone actually be a violet cone, directly responsible for our violet vision itself? The two color axes in our vision would then be violet-yellow and magenta-green. 


Edited by Jon Rista, 18 June 2023 - 01:45 PM.


#121 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 03:05 PM

As you stated, violet is a spectrally pure color. Look at the response of the blue cones...it extends into the Near UV range. As such, would blue cones themselves, not be responsible for our sensitivity to violet light, a spectrally pure color?? Why is a red response required here?

But just as with RGB filters, a stimulus of just the blue cones should produce the same sensation of blue throughout their sensitivity range, not violet, right? Their peak sensitivity is in the blue, not violet.


Edited by loujost, 18 June 2023 - 03:08 PM.


#122 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 18 June 2023 - 07:08 PM

Frank, yes, I agree with most of that. That's what this trhead is about- loss of color diversity due to mapping of all emission lines onto the same shades of R, G, and B.

 

But I can't seem to convince you of the problem with your explanation of violet in terms of bumpless sensitiviity curves. Look at your explanation carefully. It is as if you aren't really looking at the graphs.

 

"We register it as blue with a bit of red - and that translates to violet."

 

If that were the whole story, and the curves in Post #99 were correct, ask yourself where is the most red response in the blue region. It's in the region of celeste, not violet.

 

In the region where we actually see violet, the red curve shows a negligible response, approaching zero.

 

Maybe the green response cancels out the red response in the  celeste region; that does appear in some published curves. But it still seems like you need a bump in the red curve (maybe not a local maximum, but some kind of step or swelling) to make enough red response in the region where we really do see violet.

I have explained it several times but I will try again.

 

There are 3 very different types of curves shown:

 

1)  Absorbance spectra for the cones.  They show a bump in the red - but they do *not* directly correspond to what signals are generated by the cone.

 

2)  "Sensitivity" or "Response" of the cones.  They show the actual signals generated by the cones - and they do *not* show the bump

 

3)  XYZ CIE values to match RGB combinations across the spectrum.  These show a bump in X indicating that when humans combine red with blue they get a match to violet.  This is an experiment you can do for yourself and sure enough it should work

 

In order to understand how Bayer filters combined with LED displays can show purple/violet with only R, G, B light - all you need is point 3 and you can ignore 1 and 2.  You know for sure that you can create violet by combining red with blue - even though it makes no physical sense - so if you add some red sensitivity down near violet - it will all work.

 

So the only real question is why I don't see a sensitivity plot that has the bump - even though it is implied by the absorbance spectrum.  And my answer is - it doesn't need to have a bump as long as the brain just happens to turn blue with some red into violet.  The color seen doesn't need to be directly mapped to the input signals in a rigid way.

 

And there is no reason absorbance by the cone should directly map to the signal being generated.  It could simply be absorbed with no corresponding signal generated.  As long as *some* signal is generated for multiple cones in the violet region - you will see violet as violet.

 

But for entirely different reasons, it happens that when humans see blue combined with red - they see it as violet/purple.  That is why the CIE plots show a bump on the left.  That isn't a "response" curve - it's the opposite of a response curve.  It represents what a human *inputs* to drive R, G, B light sources to mimic the violet color.  It's empirical - and it's all you need to know how this stuff will work for Bayer filters to render violet with just R, G, B.

 

Frank



#123 loujost

loujost

    Mariner 2

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

Posted 18 June 2023 - 07:57 PM

I have explained it several times but I will try again.

 

There are 3 very different types of curves shown:

 

1)  Absorbance spectra for the cones.  They show a bump in the red - but they do *not* directly correspond to what signals are generated by the cone.

 

2)  "Sensitivity" or "Response" of the cones.  They show the actual signals generated by the cones - and they do *not* show the bump

 

3)  XYZ CIE values to match RGB combinations across the spectrum.  These show a bump in X indicating that when humans combine red with blue they get a match to violet.  This is an experiment you can do for yourself and sure enough it should work

 

In order to understand how Bayer filters combined with LED displays can show purple/violet with only R, G, B light - all you need is point 3 and you can ignore 1 and 2.  You know for sure that you can create violet by combining red with blue - even though it makes no physical sense - so if you add some red sensitivity down near violet - it will all work.

 

So the only real question is why I don't see a sensitivity plot that has the bump - even though it is implied by the absorbance spectrum.  And my answer is - it doesn't need to have a bump as long as the brain just happens to turn blue with some red into violet.  The color seen doesn't need to be directly mapped to the input signals in a rigid way.

 

And there is no reason absorbance by the cone should directly map to the signal being generated.  It could simply be absorbed with no corresponding signal generated.  As long as *some* signal is generated for multiple cones in the violet region - you will see violet as violet.

 

But for entirely different reasons, it happens that when humans see blue combined with red - they see it as violet/purple.  That is why the CIE plots show a bump on the left.  That isn't a "response" curve - it's the opposite of a response curve.  It represents what a human *inputs* to drive R, G, B light sources to mimic the violet color.  It's empirical - and it's all you need to know how this stuff will work for Bayer filters to render violet with just R, G, B.

 

Frank

 

"And my answer is - it doesn't need to have a bump as long as the brain just happens to turn blue with some red into violet."

 

And I have explained several times why this explanation does not work, but I'll try again. On the bumpless curves, there is more red at longer wavelengths of blue than at shorter wavelengths. Your explanation does not explain the most important fact: violet is perceived only at the shortest blue wavelengths, not the longer ones where your curve shows more red. This disproves your explanation. You keep repeating it without ever addressing this point.

 

You also say:

 

" "Sensitivity" or "Response" of the cones.  They show the actual signals generated by the cones - and they do *not* show the bump".

 

I am not an expert so I can't address this, but a reference that I found says the opposite: energy absorption of cones s directly related to signal strength. Of course, the reference may be wrong.

 

We all agree with your explanation of how RGB LEDs can display violet. That is not an issue here. Also not at issue is that non-overlapping RGB filters cannot capture and display spectrally pure violet. But the presence or necessity of a red "bump" in our cone sensitivity is I think still not resolved.

 

Regardless of the answer, in practical terms a red filter with a short-wavelength bump is required if a set of three filters is to match our visual response. And Bayer filters do have a red bump in the deep blue region.


Edited by loujost, 18 June 2023 - 07:59 PM.


#124 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 18 June 2023 - 08:55 PM

"And my answer is - it doesn't need to have a bump as long as the brain just happens to turn blue with some red into violet."

 

And I have explained several times why this explanation does not work, but I'll try again. On the bumpless curves, there is more red at longer wavelengths of blue than at shorter wavelengths. Your explanation does not explain the most important fact: violet is perceived only at the shortest blue wavelengths, not the longer ones where your curve shows more red. This disproves your explanation. You keep repeating it without ever addressing this point.

 

You are confusing "perception" of monochrome colors with "mimicing" monochrome colors using R, G, B light sources.  They are completely different.

 

We perceive violet as violet at the far end and it gets bluer as wavelength increases.  The cones could be generating anything they want - but the net perceptual response behaves like that - when simply looking at a spectrum from left to right.  Nothing magic there - cones combined with perceptual response - that's how that behaves.

 

There is no R, G, B in that scenario at all.  It is what it is and we see what we see.

 

Now switch to a totally different scenario where on the left we have a monochrome light source and on the right we have 3 separate R, G, B lamps illuminating a white piece of paper - and we set dials on those lamps to match as well as possible the pure monochrome color we see on the right.

 

For violet we dial in blue with a bit of red, and as you move to longer wavelengths you dial in less and less red.  There is a bump where red rises to a maximum - but it is combined with blue that is rising faster - and the actual proportion of red is decreasing until it rapidly goes to zero at 500nm - then it rises again.  It's the proportion in the triad that would determine the hue.

 

So I think the key thing to realize is that the "bump" in red is *not* a maximum in terms of proportion to blue.  The red proportion is steadily decreasing with wavelength if you study the graph because blue is rising much faster than red.

 

So - monochrome violet can be simulated with blue combined with a good fraction of red, and as the wavelength gets longer (less violet, more blue) the fraction of red decreases.  The fraction of red is steadily decreasing from left to right - then abruptly drops to zero - as a proportion of blue.

 

The absorbance curves are probably a decent proxy for sensitivity - but obviously they fail on the left for red - assuming the empirical curves for actual response are accurate.  One is a proxy and the other is an actual biophysical signal so I have no problem with them being different.

 

Frank



#125 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,434
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 18 June 2023 - 09:21 PM

Here is a plot of R and B values for a matched color using the tool at https://academo.org/...r-relationship/

 

RandBwavelength.png

 

And here is the ratio of R to B:

 

RtoBRatio.png

 

So there is no bump in the ratio and the fraction of R to B is strictly decreasing.

 

There may be slight variation with different conversion tools - but you can even see from the CIE plots that the fraction of red to blue is decreasing despite the peak in red.

 

So I don't see any contradictions anywhere - except that absorbance is not a useful measure for how humans perceive violet.

 

Frank


Edited by freestar8n, 18 June 2023 - 09:21 PM.



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