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

#251 BQ Octantis

BQ Octantis

    Cosmos

  • *****
  • Posts: 9,024
  • Joined: 29 Apr 2017
  • Loc: Nova, USA

Posted 27 June 2023 - 07:41 PM

What about Rec.2020?

Dumping the channels into any 3-primary gamut will map the measured RGB values to that triangle. For instance, here's the Type 2c spectrum mapped to sRGB/BT.709:

 

(Click for full size)

med_gallery_273658_12412_585151.png

 

Since the diagram is in linear space, I expect the proportions of these points along the triangle legs to be the same for all gamut triangles. But what those tones map to on the xy spectral horseshoe will depend on the gamut's primaries and white point.

 

Cheers,

 

BQ


Edited by BQ Octantis, 28 June 2023 - 06:58 PM.


#252 Jon Rista

Jon Rista

    ISS

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

Posted 27 June 2023 - 11:43 PM

"What happens if you spend 30 hours on an image, with 10 hours each in R, G and B? In contrast to say 1 hour each in R, G and B and 27 hours into L? To be perfectly honest, I don't know!!"

 

Well, you'll have a vastly greater signal in the second case, almost three times as much light, but as you say, much less color info. Might be an interesting experiment.

It would be a different signal, though... Yes, you can replace the L channel in Lab, but that is a different signal than if you actually captured deep R, G and B signals that picked up faint light deeper than say 1/10th of the signal, especially with shorter exposures, might pick up. It depends on what you are interested in. If you aren't interested in faint color, then yes, the L channel is going to pick up more signal...but indiscriminately. 

 

Part of this experiment is to see what...well, everyone (?) has been missing by NOT getting deep RGB data. Something I noticed in the first couple years after I started, was how much more faint signal light I was picking up at the dark site, that was simply impossible to pick up in a light polluted zone...even with 20+ hours, in one case about 40, I still couldn't pick up enough of these faint COLORED signals for them to show up above all the excess noise from LP. Even at the dark site, though, ~10 hours in (and that, with an OSC...my 5D III), there were still fainter signals with color, that I was just barely beginning to reveal...and that was in the relatively bright regions around say Orion, or Cygnus, etc. Head out towards Taurus, and the dark dusty region around Pleiades, and 10 hours is just barely beginning. Same thing goes for a lot of the signals that get even fainter as you move up towards the pole, where you may need even more signal, to really pick up much color at all (i.e. IFN.) A lot of the dust out there looks a faint brown and gray in most images....but...IS it really brown and gray? Or, is there more color that we just don't pick up enough of, because we don't capture enough signal with the right filters?

 

L will capture indiscriminate light faster, but you are losing something by doing so.... It is a tradeoff, not a given win. 



#253 Jon Rista

Jon Rista

    ISS

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

Posted 27 June 2023 - 11:51 PM

Dumping the channels into any 3-primary gamut will map the measured RGB values to that triangle. For instance, here's the Type 2c spectrum mapped to sRGB/BT.709:

 

(Click for full size)

med_gallery_273658_12412_259548.png

 

Since the diagram is in linear space, I expect the proportions of these points along the triangle legs to be the same for all gamut triangles. But what those tones map to on the xy spectral horseshoe will depend on the gamut's primaries and white point.

 

Cheers,

 

BQ

So this is great, because I think it shows just how challenging a color yellow actually is. REAL YELLOW, is a very narrow arc in this plot. Move just a bit out of that arc, and you are into spring green or yorange. Interestingly, I've found that reproduction of yellows with terrestrial scenes can be problematic as well. I've done a lot of nature photography, including birds and wildlife. Birds was a big deal for me...until the pandemic (i had huge plans to get back out into my bird photography that year...and it completely fell apart and I haven't done any since, which is really depressing.) There are a lot of ducks with very yellow bills...Sony's cameras actually pick up the correct yellow color best, neither Nikon nor Canon sensors can really reproduce it properly. Some ducks have yellow-green bills, others just have these amazingly yellow bills. I actually haven't tried with my EOS R5, so I can't report on that (hoping its better than the 5D III though!!), but all other DSLRs and mirrorless cameras I've tried, usually reproduce a true yellow bill as yellow green, and even with post-processing it can be really hard to get the yellow color of the bill correct. 

 

When it comes to stars...I've kind of noticed a similar issue. It is pretty easy to reproduce the reds, rusty reds, oranges, blues and whites...yellow is often the star color I see most often lacking. It frequently appears washed out, almost white, or orange. I think part of that is gapped filter sets, or LP filters that have big LP notches between reds and greens. Part of it, I think, would be a lack of proper overlap (I am not sure if the Type-2c has the most optimal overlap for it).

 

Anyway...reaching good saturation at the extremes of the spectrum (as the stated point of this thread) in the violets, reds, is troublesome, but I think yellow is also very troublesome as well, as it is really a very narrow range in the spectrum, and I suspect without a very optimal filter design with just the right overlaps, it might be a very challenging color to reproduce well in most situations (astro or terrestrial.) 



#254 BQ Octantis

BQ Octantis

    Cosmos

  • *****
  • Posts: 9,024
  • Joined: 29 Apr 2017
  • Loc: Nova, USA

Posted 28 June 2023 - 06:26 AM

Jon, in my haste to get to sleep last night, I left the plot incomplete…and I discovered this morning that it had an error on the 500 and 516.7nm vectors. I corrected and finished the plot this morning; it paints a more complete picture of the trade-offs of ramped astro filters vice color-matched Bayer filters. The updated plot is in the post above (#251).

 

BQ


Edited by BQ Octantis, 28 June 2023 - 06:27 AM.


#255 loujost

loujost

    Mariner 2

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

Posted 28 June 2023 - 06:29 AM

Yes, yellow is famously hard also in flower photography. The real point of this thread is not about problems at the extremes of the spectrum but rather about the disappearance (for spectrally pure sources) of ALL colors except R, G, and B when there are only three non-overlapping filters. But  think with yellows (and probably also purples) there may be a problem even for blackbody radiators, if the relative sensitivity of the different filters is very different from that of the human eye. Strictly speaking, though, now we are talking not about loss of color diversity but of color accuracy.


  • Jon Rista likes this

#256 Jon Rista

Jon Rista

    ISS

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

Posted 28 June 2023 - 02:11 PM

Yes, yellow is famously hard also in flower photography. The real point of this thread is not about problems at the extremes of the spectrum but rather about the disappearance (for spectrally pure sources) of ALL colors except R, G, and B when there are only three non-overlapping filters. But  think with yellows (and probably also purples) there may be a problem even for blackbody radiators, if the relative sensitivity of the different filters is very different from that of the human eye. Strictly speaking, though, now we are talking not about loss of color diversity but of color accuracy.

I think we could make an actual case study of this, with some imaging of planetary nebula with different sets of filters, overlapping, non-overlapping, gapped. For objects that DO emit narrow band spectrally pure colors, like He I and He II, and some others...Neon also emits some bands in the violet range that might be a viable option. 

 

I wonder if also capturing the same data with a range of narrow band filters for each band the nebula emits, for comparison purposes as a control case, if that might help clarify what is being lost with non-overlapping RGB filters...



#257 BQ Octantis

BQ Octantis

    Cosmos

  • *****
  • Posts: 9,024
  • Joined: 29 Apr 2017
  • Loc: Nova, USA

Posted 28 June 2023 - 09:36 PM

It dawned on me that I could do a side-by-side rendering of the by-wavelength hues from both a set of color-matched (Bayer-like) filters and a set of ramped, overlapped astronomical filters with the performance of the "Type c" filters as the exemplar. Here's how they compare:

 

(Click for full size)

med_gallery_273658_12412_114395.png

 

Color science is FUN! rainbow.gif

 

BQ


  • Borodog likes this

#258 loujost

loujost

    Mariner 2

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

Posted 28 June 2023 - 09:49 PM

It dawned on me that I could do a side-by-side rendering of the by-wavelength hues from both a set of color-matched (Bayer-like) filters and a set of ramped, overlapped astronomical filters with the performance of the "Type c" filters as the exemplar. Here's how they compare:

 

(Click for full size)

med_gallery_273658_12412_114395.png

 

Color science is FUN! rainbow.gif

 

BQ

This is great! A very reasonable result that matches our expectations. More "step-like", and losing the violets, and simplifying the reds. It is curious, though, that the 2c total spectral range is shorter than that of the Bayer-like one. Why would they clip the reds?



#259 BQ Octantis

BQ Octantis

    Cosmos

  • *****
  • Posts: 9,024
  • Joined: 29 Apr 2017
  • Loc: Nova, USA

Posted 28 June 2023 - 10:23 PM

The transition wavelengths are set by the filter ramps, but the flat regions correspond with the flat response sections of each channel with no overlap. So from left to right, it's flat blue, blue-green ramp, flat green, green-red ramp, flat red. The start and end points are set by the blue and red cutoffs.

 

As to the hue "clipping", the vertical hue cutoffs are set by dumping the filter values straight into the sRGB/BT.709 channels. That is, the flat spots are where blue, green, or red = 1, and the rest of the values are 0.

 

If you don't like those hues, then we need to "tag" the data with or transform it into the color space with the hue plateaus (i.e., primaries) we actually want.

 

BQ



#260 Jon Rista

Jon Rista

    ISS

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

Posted 29 June 2023 - 12:56 PM

This is great! A very reasonable result that matches our expectations. More "step-like", and losing the violets, and simplifying the reds. It is curious, though, that the 2c total spectral range is shorter than that of the Bayer-like one. Why would they clip the reds?

Great indeed! 

 

I do wonder about the viability of deeper violet capture... A lot of scopes, refractors, will scatter much more of the deeper blues, violets and UV, which can create some notable effects around stars that are bright enough. Some filter manufacturers explicitly try to avoid passing too deep of blues/violets, to avoid that kind of scatter. I know my FSQ, and I know some other scopes I looked very seriously at, can have some significant blue scattering issues. 

 

So even if the Type-2c passed those deeper wavelengths, what would you actually be capturing in the end...something coherent and structured, or just a scattered mess? 


Edited by Jon Rista, 29 June 2023 - 02:29 PM.


#261 loujost

loujost

    Mariner 2

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

Posted 29 June 2023 - 01:05 PM

Great indeed! 

 

I do wonder about the viability of deeper violet capture... A lot of scopes, reflectors, will scatter much more of the deeper blues, violets and UV, which can create some notable effects around stars that are bright enough. Some filter manufacturers explicitly try to avoid passing too deep of blues/violets, to avoid that kind of scatter. I know my FSQ, and I know some other scopes I looked very seriously at, can have some significant blue scattering issues. 

 

So even if the Type-2c passed those deeper wavelengths, what would you actually be capturing in the end...something coherent and structured, or just a scattered mess? 

That's a good point, and not something that color science could answer, since it depends on the quality of the optics, though I thought this would be more of a problem with refractors (because of chromatic aberrations). As I'm sure you know, many filter makers even sell  "minus violet" filters to reduce star bloat.



#262 Jon Rista

Jon Rista

    ISS

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

Posted 29 June 2023 - 02:29 PM

That's a good point, and not something that color science could answer, since it depends on the quality of the optics, though I thought this would be more of a problem with refractors (because of chromatic aberrations). As I'm sure you know, many filter makers even sell  "minus violet" filters to reduce star bloat.

Ack, sorry...I meant refractors... I even mentioned my own refractor. :p 



#263 slippy

slippy

    Explorer 1

  • *****
  • Posts: 66
  • Joined: 10 Sep 2021
  • Loc: Seattle

Posted 15 April 2024 - 01:09 AM

This seemed like an obvious shortcoming when I saw the filter cutoffs, and has bothered me for years. Other imagers tried to convince me that I just misunderstood color science. Haha, a bit frustrating to read so many people not getting the point, even with the image in the first post, or trying to handwave it away as not necessary. Props to Lou for persisting.

 

There are known signals that cannot be differentiated with the common filter sets. That's the point! It matters less that they perfectly match human perception IMO, but at the very least, there should be somewhere to map them, and the human eye response is proof that there is at least a way to triangulate them in the color space that we can prove works where RGB doesn't.

 

Also, I think aside from the filter curves you'd also have to factor in the QE curve for your camera. That throws everything off. But a calibration should be possible.

 

BTW, I believe fireworks are another example. Can't RGB some of those.

 

Jon, did you ever try imaging with your other filter?



#264 loujost

loujost

    Mariner 2

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

Posted 17 April 2024 - 12:43 AM

Thanks for the kind words and a very good summary of the issue, Explorer 1. It can be exhausting to deal with the negative comments, here and in my other posts, but with patience, eventually the discussion becomes productive.


  • Domdron likes this

#265 slippy

slippy

    Explorer 1

  • *****
  • Posts: 66
  • Joined: 10 Sep 2021
  • Loc: Seattle

Posted 17 April 2024 - 03:59 AM

Looks like chroma does have cie tristimulus filters. Weak in the red for astro though.

 

https://www.chroma.c...ulus-chart.png tristimulus-chart.png



#266 loujost

loujost

    Mariner 2

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

Posted 17 April 2024 - 07:57 AM

Wow, great find! This is a game-changer. We could also add an Halpha filter to teak the result. But I get a Page not Found message at that link,


Edited by loujost, 17 April 2024 - 07:58 AM.


#267 Borodog

Borodog

    Voyager 1

  • *****
  • Posts: 10,933
  • Joined: 26 Oct 2020
  • Loc: St. Augustine, FL

Posted 17 April 2024 - 08:31 AM

Everything in AP, other than imaging the Moon sometimes, is about making images that are *not* how the human eye sees things.
  • licho52 and Robert7980 like this

#268 slippy

slippy

    Explorer 1

  • *****
  • Posts: 66
  • Joined: 10 Sep 2021
  • Loc: Seattle

Posted 17 April 2024 - 09:33 AM

And another who completely misunderstands the issue…



#269 slippy

slippy

    Explorer 1

  • *****
  • Posts: 66
  • Joined: 10 Sep 2021
  • Loc: Seattle

Posted 17 April 2024 - 09:35 AM

Wow, great find! This is a game-changer. We could also add an Halpha filter to teak the result. But I get a Page not Found message at that link,

https://www.chroma.c...imulus-filters 



#270 badgie

badgie

    Viking 1

  • *****
  • Posts: 637
  • Joined: 13 Jan 2021
  • Loc: SF Area, California

Posted 22 April 2024 - 05:55 PM

If you want to look at this link, remove the  "%A0" at the end, clicking on the link above fails for me. 

 

Hopefully this one works



#271 martz

martz

    Lift Off

  • -----
  • Posts: 24
  • Joined: 26 Aug 2020

Posted 23 April 2024 - 07:29 AM

Wow, great find! This is a game-changer. We could also add an Halpha filter to teak the result. But I get a Page not Found message at that link,

Question:  What would be the conceptual advantages of monochrome camera imaging with filters such as the chroma tristimulus filters (assuming they are apt for photography), other than obtaining increased data per color per time (not an insignificant consideration), versus OSC/DSLR imaging, with a combined imaging system (i.e., sensing materials, sensor lenses, UV/IR filters, CFAs, algorithms, etc.) that a manufacturer natively aims to optimize to approximate human color perception in daylight?

 

This is the experienced imagers' forum and am not one, so please forgive my ignorance, but this topic interests me.  Thank you.    



#272 loujost

loujost

    Mariner 2

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

Posted 23 April 2024 - 08:02 AM

Question:  What would be the conceptual advantages of monochrome camera imaging with filters such as the chroma tristimulus filters (assuming they are apt for photography), other than obtaining increased data per color per time (not an insignificant consideration), versus OSC/DSLR imaging, with a combined imaging system (i.e., sensing materials, sensor lenses, UV/IR filters, CFAs, algorithms, etc.) that a manufacturer natively aims to optimize to approximate human color perception in daylight?

 

This is the experienced imagers' forum and am not one, so please forgive my ignorance, but this topic interests me.  Thank you.    

That's a good question. Even the advantage you mentioned (increased data per color per unit time) is very small. There is an advantage in the red and blue bands and a disadvantage in the green band. because of the way Bayer filters are designed. The only way to get a clear advantage is to spend some of the time taking luminance data. And that can be important.

 

But the advantage of being able to choose the filters means that you can, if desired, boost the H_alpha signal, which is weakly captured by most DSLRs. In addition, astro cameras (both mono and OSC) usually have cooled sensors, giving them a big advantage over DSLRs and other cameras deigned for general use.

 

For me a big advantage is that we could now use just one cooled-sensor camera for all types of astrophotography: narrow-band imagery, natural color, H_alpha-enhanced natural color, UV, IR etc.

 

Lou


Edited by loujost, 23 April 2024 - 08:02 AM.

  • martz likes this

#273 badgie

badgie

    Viking 1

  • *****
  • Posts: 637
  • Joined: 13 Jan 2021
  • Loc: SF Area, California

Posted 23 April 2024 - 12:14 PM

I don't think I would want to use the tristimulous for any "normal" narrowband objects, even though they would have great color rendition.  Like our eyes, there is poor sensitivity for both OIII and Ha.  (~20% T)  That is a killer for me on any object where I would care about those emission lines. 



#274 slippy

slippy

    Explorer 1

  • *****
  • Posts: 66
  • Joined: 10 Sep 2021
  • Loc: Seattle

Posted 23 April 2024 - 12:53 PM

If you want to look at this link, remove the  "%A0" at the end, clicking on the link above fails for me. 

 

Hopefully this one works

 

Thanks, somehow I messed that up twice!

 

 

Anyway, as for the advantage, OSC probably already does a decent job of this. However, it’s inherently lower resolution due to the bayer matrix, and since most people already have and prefer mono cams, it would be a way to use those similar to how they use them for RGB data today, just with a bit more spectrum coverage. But because they’re all curves, it would still be less efficient than RGB.

 

I’m still not sure I’d bother going the tristimulus route anytime soon, just exploring ideas.



#275 BQ Octantis

BQ Octantis

    Cosmos

  • *****
  • Posts: 9,024
  • Joined: 29 Apr 2017
  • Loc: Nova, USA

Posted 23 April 2024 - 04:29 PM

For me a big advantage is that we could now use just one cooled-sensor camera for all types of astrophotography: narrow-band imagery, natural color, H_alpha-enhanced natural color, UV, IR etc.

You can already do this with an OSC. We recently proved we could color match with two astrocams—the ASI224MC and the ASI533MC—to very low E. All you need is a UV/IR cut filter. We even got the True Colors of the planets using the lunar regolith to model and remove red shift from atmospheric absorption…

 

https://www.cloudyni...th-an-asi224mc/

 

Take your astrocam, plug in the UV/IR cut, shoot a CC24 under D65 illumination, and Bob's your uncle.

 

BQ


Edited by BQ Octantis, 23 April 2024 - 06:30 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