Using the moon as a grey card for colour balance?

[Tulloch]

Hi all, here's one that's a bit left field for you.

I do astrophotography in my spare time, and most dedicated colour astro-cameras do not perform any colour balancing at all, instead they simply record the light through the Bayer filter on the sensor and send it to a computer in linear format. No WB corrections, no icm colour matrix, no gamma corrections, just raw, linear numbers which we then interpret as a colour image.

While it is possible to point the astro-camera at a Macbeth colour chart and produce a colour calibration matrix, this assumes a particular colour temperature which is probably not correct for night-time imaging (ie even though the light from the sun passing through the atmosphere is dispersed in the same way, the sky is not blue at night and so doesn't contribute to the temperature of the light source).

One possible way to calibrate these cameras is by using the moon as a pseudo-grey card target. While the moon isn't completely colourless (it is slightly more red than blue - see the attached graph) and reflects about 12% of the light from its surface, it's about the closest thing I can think of.

What do people think of this as an idea? What are the problems with this approach? Are there any better ways of calibrating a linear camera sensor for night-time imaging that I haven't thought of?

Thanks in advance,

Andrew

[xpatUSA]

Hi all, here's one that's a bit left field for you.

I do astrophotography in my spare time, and most dedicated colour astro-cameras do not perform any colour balancing at all, instead they simply record the light through the Bayer filter on the sensor and send it to a computer in linear format. No WB corrections, no icm colour matrix, no gamma corrections, just raw, linear numbers which we then interpret as a colour image.

While it is possible to point the astro-camera at a Macbeth colour chart and produce a colour calibration matrix, this assumes a particular colour temperature which is probably not correct for night-time imaging (ie even though the light from the sun passing through the atmosphere is dispersed in the same way, the sky is not blue at night and so doesn't contribute to the temperature of the light source).

One possible way to calibrate these cameras is by using the moon as a pseudo-grey card target. While the moon isn't completely colourless (it is slightly more red than blue - see the attached graph) and reflects about 12% of the light from its surface, it's about the closest thing I can think of.

What do people think of this as an idea? What are the problems with this approach? Are there any better ways of calibrating a linear camera sensor for night-time imaging that I haven't thought of?

Thanks in advance,

Andrew

One possible problem is that the graph is for radiometric units and color is based on photometric units.

"While the moon isn't completely colourless" ...

Another is that the moon's color is quite variable - for example blood or harvest moon, so I'm wondering what range of error is acceptable in your work?

Drifting off-topic, I don't shoot the moon much and certainly no astro stuff so I often convert the image to grayscale and then color to taste:

[Tulloch]

One possible problem is that the graph is for radiometric units and color is based on photometric units.

"While the moon isn't completely colourless" ...

Another is that the moon's color is quite variable - for example blood or harvest moon, so I'm wondering what range of error is acceptable in your work?

I understand (especially recently with the lunar eclipse just past), but I was hoping to shoot it in a month or so when it is almost directly overhead to reduce any effects of the atmosphere. I guess I'm simply asking how I could use this in my images - is it just a matter of measuring the average r,g,b values across the disc and comparing it to the expected spectra of the moon to provide a correction factor for the red and blue channels compared to green?

How do DSLRs do it?

[xpatUSA]

I guess I'm simply asking how I could use this in my images - is it just a matter of measuring the average r,g,b values across the disc and comparing it to the expected spectra of the moon to provide a correction factor for the red and blue channels compared to green?

No.

Hello again.

In post, my editor (RawTherapee) has a white balance function where the cursor is a circle selectable up to 32px diameter. So you could shoot the moon and click on it in several places - taking note of the resulting degrees K and of the tint values. Average them and then set the white balance sliders to that average. Then use those same average values for other shots in the same session.

Maybe your editor has a similar function?

How do DSLRs do it?

Most DSLRs have a Custom White Balance function - but it is necessary on mine to fill the frame with the presumably neutral content to get a reasonable WB.

HTH.

[Manfred M]

Unless you can get someone to place a giant white balance card on the moon for you, your idea is not going to work. These cards are used by placing them so that they reflect light from the same source as it hitting the subject, and the white balance / colour temperature is calculated based on the values that we read from the target and then corrected against the actual known values of the card (often, as you point out, RGB values).

When we discuss colour in photography, we have to look at three separate issues:

1. Physics; i.e. what is the equivalent black body radiator temperature of the light source is. This is effectively what a spectrally neutral colour target gives us. Impacts like Rayleigh scattering (hence the warmer colours of the moon when it gets closer to the horizon) is another impact of physics;

2. Human physiology - how the human visual system works (i.e. the eye to brain) processing of the image. The human visual system responds very quickly to changes in colour vision; we can step inside on a brightly lit winter day with north light (colour temperature well above 10000K) and into a darkened room with a roaring fire (colour temperature at below 2000K), yet both will look "right" when we see things; the eye starts to adapt to different colour temperature light sources in a fraction of a second (colour adaptation).

There are other physiological effects associated with low light. The vast majority of the light receptors in the human eye are "rods" that are not colour sensitive, but the "cone" shaped ones is how we perceive colour. Both types of receptors influence how we perceive colours and in darker conditions, like when you are photographing the moon, the Purkinje effect comes into play and creates a blue shift in our colour perception.

3. Psychological impact of colour - perhaps one that is less important in this aspect of photography, but deals with how we interpret colours. Warm tones and cool tones are one such outcome when we look at the psychological impact of colours. Expecting to see a "silver" moon when it is high in the sky or a warmer toned moon as it drops close to the horizon are two such impacts. If you show a warm toned moon when it is at the zenith or a cool tone when the moon is at the horizon, people are going to feel that there is something wrong with the image.

The bottom line is that there is no correct colour temperature for what you are trying to do, although I have read somewhere that 4000K is probably close to being a "correct" value.

How do DSLRs (and other cameras) do it? They have algorithms that calculate an appropriate colour temperature to use, but they have their limitations. When I look at my camera manual tells me that its algorithms works for colour temperatures between 3500K and 8000K, otherwise I have to set the colour temperatures using another method. As a raw shooter (much like how you describe your astrophotography camera), I set my colour temperatures with my raw convertor and depending on the scene, I use different approaches for doing so.

The automatic algorithms in these cameras can be fooled into giving poor colour renditions. Mixed light sources (i.e. different colour temperature sources) will fool the algorithms as will scenes that have other external influences (light filtering through tree leaves will result in a green colour cast, photograph someone standing beside a red brick wall, and they will have a red colour cast, etc.

I hope that this helps...

[xpatUSA]

Unless you can get someone to place a giant white balance card on the moon for you, your idea is not going to work. <big snip>

Hopefully the OP will not be put off by this and the large amount of related advice.

Meanwhile, some of these might help:

https://www.dpreview.com/forums/thread/3970742

https://blogs.scientificamerican.com...ooks-that-way/

https://newbedev.com/why-does-moonli...or-temperature
.

[Manfred M]

Hopefully the OP will not be put off by this and the large amount of related advice.

The "large amount of advice" boils down to that there is no correct answer. Do what works for you.

Just as an aside, the following is an image of the Perseverance Mars Rover MastCam-Z calibration target.

[xpatUSA]

The "large amount of advice" boils down to that there is no correct answer.

Do what works for you.

I'm not much of a moon-shooter myself. I haven't got the gear.

I'm still a bit puzzled by the thread title:

"Using the moon as a grey card for colour balance?"

If that means using a portion of the moon in order to shoot the moon itself at zenith and nothing else, then research indicates that the dirt is good enough for the purpose and should render the slight color of the mares reasonably well.

If, on the other hand, it means shooting the moon as if it were a gray card for the purpose of color balancing everything else in the scene then everything you said applies including that it won't work.

Just as an aside, the following is an image of the Perseverance Mars Rover MastCam-Z calibration target.

Excellent! Also quite cute ...

[Tulloch]

No.

Hello again.

In post, my editor (RawTherapee) has a white balance function where the cursor is a circle selectable up to 32px diameter. So you could shoot the moon and click on it in several places - taking note of the resulting degrees K and of the tint values. Average them and then set the white balance sliders to that average. Then use those same average values for other shots in the same session.

Maybe your editor has a similar function?

HTH.

Hi again Ted, thanks for your replies. I have been using RawTherapee in the past to modify the colour profiles I've created using my Astro-camera and Macbeth colour charts. I've found that it's an extremely powerful tool, with many features I've never heard of, and requires a significant degree of reading to understand what it's capable of. However, for simply measuring the colour temperature at different locations it could be useful .

[EDIT:] OK, so I had a play in RawTherapee on an image of the moon I took at the end of the recent lunar eclipse. The moon was full, sitting at about 67* elevation, and I photographed it using my Canon 700D using Daylight WB (since I have to use something) through a telescope. Selecting the brighter parts of the moon came up with a recommended colour temperature of 6200K, the dark areas came up at 6000K. These values were pretty consistent.

What does this mean? I imaged the moon using a WB value of 5200K, and RT thinks it should be 6100K.

[Tulloch]

Unless you can get someone to place a giant white balance card on the moon for you, your idea is not going to work. These cards are used by placing them so that they reflect light from the same source as it hitting the subject, and the white balance / colour temperature is calculated based on the values that we read from the target and then corrected against the actual known values of the card (often, as you point out, RGB values).

When we discuss colour in photography, we have to look at three separate issues:

1. Physics; i.e. what is the equivalent black body radiator temperature of the light source is.

2. Human physiology - how the human visual system works (i.e. the eye to brain) processing of the image.

3. Psychological impact of colour - perhaps one that is less important in this aspect of photography, but deals with how we interpret colours.

The bottom line is that there is no correct colour temperature for what you are trying to do, although I have read somewhere that 4000K is probably close to being a "correct" value.

How do DSLRs (and other cameras) do it? They have algorithms that calculate an appropriate colour temperature to use, but they have their limitations. When I look at my camera manual tells me that its algorithms works for colour temperatures between 3500K and 8000K, otherwise I have to set the colour temperatures using another method. As a raw shooter (much like how you describe your astrophotography camera), I set my colour temperatures with my raw convertor and depending on the scene, I use different approaches for doing so.

I hope that this helps...

Hi Manfred, that helps a lot . In regards to your specific points:

1. Since I am imaging the planets, the light source (the Sun) should be the same (or at least similar) for both the Moon and (say) Jupiter, neglecting any effects of interstellar dust. Therefore, the colour temperature at the target (Jupiter) should be the same as that at the Moon. What is different is that the intensity of the light hitting Jupiter is a lot less than the Moon, and I'm unsure if that makes a difference to any WB calculations.

2. Looking through a largish telescope, the close planets are very bright (especially Mars, Jupiter and Saturn), so we are certainly in the photopic region of the eye response (which of course does not affect cameras), while the moon is almost blinding. I (and others) have measured the colour shift with elevation due to the atmosphere, and it's not until the planets are lower than around 45 degrees in elevation that the shift is significant (maybe a few percent in the red/blue channel compared to green).

3. I have gone through the CIE 1931 standard for human vision, and have converted the measured spectral response by others into XYZ and hence RGB (assuming sRGB) and obtained an "average" colour for the planets, but it's difficult to convert that into a colour correction tool for my images.

I have seen a number of websites that recommend the correct WB values to use, and am also coming around to a lower temperature (around 4000K). I believe that the "Daylight" setting (that others recommend) is removing too much of the blue channel, which is present in terrestrial photography but missing in night-time astrophotography.

I would really appreciate your insight into my statements above, especially if my thinking/assumptions are just plain wrong in any way. I have been looking into this for a while now (you can read about some of my exploits here if you have a few hours to kill ), so I would greatly appreciate your thoughts as a professional photographer.

Thanks again,

Andrew

[Tulloch]

The "large amount of advice" boils down to that there is no correct answer. Do what works for you.

Just as an aside, the following is an image of the Perseverance Mars Rover MastCam-Z calibration target.

Glad they brought one, hopefully it won't get covered in red dust

[xpatUSA]

OK, so I had a play in RawTherapee on an image of the moon I took at the end of the recent lunar eclipse. The moon was full, sitting at about 67* elevation, and I photographed it using my Canon 700D using Daylight WB (since I have to use something) through a telescope. Selecting the brighter parts of the moon came up with a recommended colour temperature of 6200K, the dark areas came up at 6000K. These values were pretty consistent.

What does this mean? I imaged the moon using a WB value of 5200K, and RT thinks it should be 6100K.

RT can only see what your capture and conversion did.

RT also has a CIELAB color picker which I used on your posted image:

x y a* b*

337 183 0.5 3.3
236 210 -0.2 2.8
378 294 -0.2 2.6

These numbers are fairly close to gray but with a tendency toward yellow/orange. In Kelvin terms, a tendency to "warm". So the Canon was not set warm enough if the goal was to capture the moon as perfectly gray (a*,b* = 0,0). Therefore, RT's WB picker showed you the "correct" WB to achieve that.

Here's your shot with the hues extracted in the GIMP:



The hues are show at 100% saturation in this view. Remember, hue exists even at almost no saturation - so the bias towards yellow/orange is quite clear in the above.

The red background comes from the GIMP which interprets a grayscale color (R=G=B) as a hue of red, i.e. 0 or 360 degrees of hue.

Off-topic but there's a nice red moon here:

https://www.dpreview.com/forums/post/65183838
.

[xpatUSA]

... What is different is that the intensity** of the light hitting Jupiter is a lot less than the Moon, and I'm unsure if that makes a difference to any WB calculations.

It does not, in theory, Andrew. But the 1931 xyY diagram is a projected slice from a 3D volume so, if a particular WB setting goes out-of-gamut, then that could give rise to an anomaly.

** "Intensity" is not quite the right word. "Illuminance" is better because intensity is more often used for the brightness of a light point source independent of distance.

[Tulloch]

RT can only see what your capture and conversion did.

RT also has a CIELAB color picker which I used on your posted image:

x y a* b*

337 183 0.5 3.3
236 210 -0.2 2.8
378 294 -0.2 2.6

These numbers are fairly close to gray but with a tendency toward yellow/orange. In Kelvin terms, a tendency to "warm". So the Canon was not set warm enough if the goal was to capture the moon as perfectly gray (a*,b* = 0,0). Therefore, RT's WB picker showed you the "correct" WB to achieve that.

Off-topic but there's a nice red moon here:

Thanks again Ted, very helpful . It wasn't my intention to try and capture a perfectly grey moon, rather since I was capturing in jpg format (rather than raw due to file size and stacking considerations), I just chose "Daylight" WB on my DSLR.

I took a few images of the moon during the recent lunar eclipse (of which the image above was last in the sequence), which you can see here (scroll down to see the 12 frame panel).
https://www.cloudynights.com/topic/7...a-26-may-2021/

[xpatUSA]

Thanks again Ted, very helpful . It wasn't my intention to try and capture a perfectly grey moon, rather since I was capturing in jpg format (rather than raw due to file size and stacking considerations), I just chose "Daylight" WB on my DSLR.

I see. That explains the yellow/orange cast, I reckon.

So, may I ask, what was the intention in terms of color for the final processed images?

I took a few images of the moon during the recent lunar eclipse (of which the image above was last in the sequence) ... <>

Thanks for the link, here's one of your thumbnails:



An interesting difference between the middle row and the others ...

... I blame the Canon

[Tulloch]

Well, you should blame me, as we entered totality I increased the ISO from 100 to 800, and increased the exposure from 1/1000 to 0.6 sec to show up the red

[Manfred M]

Hi Manfred, that helps a lot . In regards to your specific points:

1. Since I am imaging the planets, the light source (the Sun) should be the same (or at least similar) for both the Moon and (say) Jupiter, neglecting any effects of interstellar dust. Therefore, the colour temperature at the target (Jupiter) should be the same as that at the Moon. What is different is that the intensity of the light hitting Jupiter is a lot less than the Moon, and I'm unsure if that makes a difference to any WB calculations.

Sorry for the delay in getting back, I've been busy for the past few days.

First of all, you are right about the light hitting these celestial bodies; the light source is the same, but the intensity varies. The inverse square law on a very large scale.

What you are forgetting about are the impacts of the earth's atmosphere and that effectively acts as if you have put a somewhat fuzzy filter in front of your telescope opening.Ultimately, these effects are what are going to impact your final images; primarily the Rayleigh scattering. Some of this can be reduced by shooting close to vertically, but frankly some of the worst impacts will occur closer to the ground. Atmospheric moisture, particulate matter from human activity and volcanic activity or forest fires all cause issues.

2. Looking through a largish telescope, the close planets are very bright (especially Mars, Jupiter and Saturn), so we are certainly in the photopic region of the eye response (which of course does not affect cameras), while the moon is almost blinding. I (and others) have measured the colour shift with elevation due to the atmosphere, and it's not until the planets are lower than around 45 degrees in elevation that the shift is significant (maybe a few percent in the red/blue channel compared to green).

Photonic effects are really only going to impact images of the moon. People "know" what the moon looks like, so their memory is going to look like as we see it often, so colour correcting it could actually be counter-productive as it might not match what we expect. The "blood moon" gets our attention because the moon does not look "right".

3. I have gone through the CIE 1931 standard for human vision, and have converted the measured spectral response by others into XYZ and hence RGB (assuming sRGB) and obtained an "average" colour for the planets, but it's difficult to convert that into a colour correction tool for my images.

I agree and I'm not quite sure what you are trying to achieve here. Unless the planets are actually physically close to having a neutral colour, all you are going to end up doing is creating a "false" colour. Take the earth, with all its water; it is hardly going to give you a neutral white balance and if you do, the image is going to end up looking quite warm as the complementary colours of blue, i.e. yellow and complement of cyan, i.e. red will dominate the image. Mars, Jupiter and Saturn all do not look like they are neutral coloured. As you have noted, the moon is a bit more warm toned, but the "silver" colour we normally associate with the moon is really due to the light coming through the atmosphere.

I have seen a number of websites that recommend the correct WB values to use, and am also coming around to a lower temperature (around 4000K). I believe that the "Daylight" setting (that others recommend) is removing too much of the blue channel, which is present in terrestrial photography but missing in night-time astrophotography.

This inherently makes sense to me and we often see this in genre; images with daylight white balance come out looking a bit too warm and cutting back to a lower colour temperature likely results in more pleasing looking images.

In "earth-bound" photography, we can get into all kinds of discussions on what the "correct" colour temperature should be to get a neutral white balance. In my view, the correct answer is generally what the the "client" is looking for. If that means a white balance selected based on the light reflected from a spectrally neutral white balance target, so be it. If the client wants a really weird colour grade, so be it. For us amateurs, we are generally our own client. I know in my own work, I don't always white balance to neutral, especially in portraiture work.

In astrophotography, the idea of a "correct" white balance is more of a philosophical position than purely a technical one. The photographer is taking a subject that is (with the exception of the sun or moon) something that no one has seen. The images that we have been shown are ones that have been photographed by telescopes, especially with space based platforms. The only reference we have are published images and there is no right or wrong here.

[xpatUSA]

Well, you should blame me, as we entered totality I increased the ISO from 100 to 800, and increased the exposure from 1/1000 to 0.6 sec to show up the red

So the intention in terms of color for the final processed images was neutral or red, depending on the phase of the eclipse.

[Tulloch]

What you are forgetting about are the impacts of the earth's atmosphere and that effectively acts as if you have put a somewhat fuzzy filter in front of your telescope opening. Ultimately, these effects are what are going to impact your final images; primarily the Rayleigh scattering. Some of this can be reduced by shooting close to vertically, but frankly some of the worst impacts will occur closer to the ground. Atmospheric moisture, particulate matter from human activity and volcanic activity or forest fires all cause issues.

The effects of Rayleigh scattering, ozone and aerosols can be modelled and the effects removed. This is dependent upon the elevation of the target, and can be account for. It should be same for the moon and the planets, assuming the elevation is the same.
http://astrosurf.com/aras/extinction/calcul.htm

Photonic effects are really only going to impact images of the moon. People "know" what the moon looks like, so their memory is going to look like as we see it often, so colour correcting it could actually be counter-productive as it might not match what we expect. The "blood moon" gets our attention because the moon does not look "right".

It doesn't matter what people think the moon looks like, the spectrum can (and has) been measured over the visible wavelengths.

I agree and I'm not quite sure what you are trying to achieve here. Unless the planets are actually physically close to having a neutral colour, all you are going to end up doing is creating a "false" colour. Take the earth, with all its water; it is hardly going to give you a neutral white balance and if you do, the image is going to end up looking quite warm as the complementary colours of blue, i.e. yellow and complement of cyan, i.e. red will dominate the image. Mars, Jupiter and Saturn all do not look like they are neutral coloured. As you have noted, the moon is a bit more warm toned, but the "silver" colour we normally associate with the moon is really due to the light coming through the atmosphere.

I'm not trying to create a neutral white balance for the planets - I'm trying to use a (more or less) neutral target - the Moon - as the "giant grey card in the sky" to balance my camera so that the colour of the targets - the planets - are accurately represented.

In astrophotography, the idea of a "correct" white balance is more of a philosophical position than purely a technical one. The photographer is taking a subject that is (with the exception of the sun or moon) something that no one has seen. The images that we have been shown are ones that have been photographed by telescopes, especially with space based platforms. The only reference we have are published images and there is no right or wrong here.

The "published images" are normally modified for maximum impact with the public. Take the images from Hubble for instance, the camera filters used by Hubble are very narrow band and designed for bringing up the colours of distant nebulae and galaxies, not accurately colour images of the planets. Sure, it can be used for that, but the results are not "accurate".

There should be no reason why astro-images cannot be represented with colour accuracy, any more than any terrestrial scene can be represented with colour accuracy. Obviously it is more difficult since we don't have a giant grey card in the sky, but I don't believe there should be any reason why the moon could not be used in this way, with an appropriate modification since it's not perfectly grey. Just wondering how to do it

[Tulloch]

So the intention in terms of color for the final processed images was neutral or red, depending on the phase of the eclipse.

The intention was to try and capture the eclipse, and as the eclipse proceeded I needed to increase the exposure level so the camera captured enough light for an image. No colour modification was performed post capture (aside from some increase in saturation levels for the eclipsed moon).