Hi everyone! I need your expert advice. I am attempting to print a night shot for the first time and I'm worried about the image quality. Above is a photo from a remote corner of Rocky Mountain National Park (Erik, you'll know where this is ) taken last Sunday. I've gotten a few requests for prints but I am unsure of how to prepare the image for best results.
First of all, my blend isn't the cleanest - so if you have suggestions I'd love to hear them. Second, the sky shot was taken at ISO 6400 and is thus quite noisy. How can I avoid producing an overly grainy photo? I've done plenty of noise reduction in Lightroom but still see quite a bit of grain. I will not be printing very large - 12x18 at the most with the remote possibility of 16x24.
Image stats:
Nikon D5100
Tokina 11-16mm
11 mm
foreground - ISO 3200 approximately 8 minutes
sky - ISO 6400 20 seconds
f/2.8
Thank you all SO much!
Posted by Roger Clark on August 26, 2014 at 11:22 AM
Hello Crystal,
Beautiful image and at least not too extreme on the color balance (though popular these days to have blue stars on a blue sky--note there really are no blue stars and the Milky Way is not blue). You do have some nice reddish brown in the Milky way. The star Anatares is reddish-orange and does appear orange to the eye, but is white/blueish-white in your image. Similar with other stars: colors can be seen in brighter stars, more with binoculars.
Lots to discuss here. Not sure what you did in post processing, but ISO 6400 didn't help--most of the stars are saturated. ISO does not change sensitivity. So changing ISO will not change the noise, only exposure time and lens aperture will help that. I have a new article on post processing nightscapes to maintain star color:
http://www.clarkvision.com/articles/night.photography.image.processing/
For two exposures, land and sky, I keep them on separate layers. That can help avoid the bright rim around the mountain tops when trying to work
the image of the sky versus land.
Roger
Roger
Posted by Jack Brauer on August 26, 2014 at 1:26 PM
Crystal - That is a wonderful photo! My advice regarding the printing would be to do a test print first; you might find that it ends up looking significantly dimmer in print (at least under normal lighting conditions) than on a bright backlit monitor, and may need brightening beyond what might look best on screen.
Roger - Your articles just blew my mind! Very enlightening (no pun intended) information and explanations of ISO. It's amazing, and humbling, that after over a decade of digital photography I only now realize that I misunderstood ISO. Quite enlightening to learn that boosting the ISO past a certain point is not doing anything helpful to the image, and that it is better done in post processing after that point. I spent some time testing this in Lightroom with various night shots at same settings but different ISOs, and sure enough your explanations hold true. So thank you for that. Your website has a wealth of useful information.
One thing I would say - and I'm sure you've heard this a thousand times - is that the bluer night sky photos are perhaps closer to our perceptual reality. I understand and respect your statements about true sky/star colors, but also no human eye can see the galaxies and stars so clear as a camera can so we're already in "unreal" territory at this point. I would argue that the (slightly) bluer treatment is or can be closer to our subjective/perceptual reality of star gazing than a redder (though scientifically truer) rendition of the night sky.
The same idea would apply to moonlight photos (though I realize this is a different topic and more extreme example). When properly exposed, a moonlit scene might look just like daylight. But by processing the photo much bluer (for a "whiter" light rendition), the photo will more accurately convey the experience of how we see in the moonlight.
Posted by Jack Brauer on August 26, 2014 at 3:51 PM
@Roger again (sorry Crystal, I don't mean to hijack your thread!) - Having spent even more time reading your articles, and also viewing your fantastic night photos, I should clarify that my statements are aimed more towards casual/occasional night shooters such as myself whose photos are not usually capturing so much of the Milky Way and other atmospheric phenomenon but rather simply a landscape under a sky full of stars. What you are doing is clearly on another level! Here's a perfect example of what I mean: http://www.mountainphotography.com/photo/lunar-alpenglow/?search=moonlight In that photo I purposely shifted the sky colors from their truer browner shades to a richer, bluer color.... mainly just personal color contrast preference in this case, but also because I do think in this instance it looks more "real" in the sense that people (who are not astrophotographers) can relate more to a bluer night sky than a brown one.
Your article here ( http://www.clarkvision.com/articles/color.of.the.night.sky/ ) is a great read and explains this issue thoroughly. Of particular interest related to my own statements would be your examples in Figures 10a and 10b. 10b is clearly the most beautiful and as you say accurate color rendering; however, were this my photo, I would be very much tempted to shift the background brown color (in the blacks) ever so slightly towards blue (not nearly as dramatically as 10a) which in my own opinion would appear more realistic to what my eyes would see (even if it's not).
It's a very interesting topic to me (thus my lengthy comments here!), and a topic you've tackled more thoroughly than any other I've run across.
Posted by Roger Clark on August 26, 2014 at 6:04 PM
Hi Jack,
You say "One thing I would say - and I'm sure you've heard this a thousand times - is that the bluer night sky photos are perhaps closer to our perceptual reality."
Well, I'm not sure a thousand times, but yes i have heard it a few times. But it is a recent, last few years fad, brought on in the digital era. And I would say the perception is highly colored (pun intended as you will see).
Let me go back to your lunar alpenglow image: beautiful!!!! (perhaps just a little too blue ;-) When the Moon is up in the sky and more than just a thin crescent, we do see Rayleigh scattering and a blue night sky.
But with no Moon and away from city lights, if one wanted to color their images the way they are perceived, it should be gray, not blue. I have studied the night sky for over 50 years. In the last year because of these issues, I've studied what colors can I see in the dark night sky. My conclusions so far as that people are seeing blue probably because of the bright red lights they are using. A strong color will give a strong opposite color response. There is another myth that red light won't hurt night vision. Yes they do, and it warps color perception too.
Try this experiment next time you are out under a dark, moonless night sky. Use no lights of any kind for 30 minutes. The sky will be mostly gray unless there is really strong airglow or aurora. Now, is the gray a warm gray, or a cool gray? When I do this, I become very well dark adapted with no light for 30 minutes, an hour better, I can see the color tint of the gray. Many stars show color, from blueish-white to yellow, orange and red. The core of the Milky Way in Sagittarius appears brownish gray. I can see banded airglow and can pick out the bands that are redder and those that are greener. If there is some auroral glow in the north, even when faint, I can pick out the color. A quick (10 to 30 second) exposure on a digital camera shows me if I am right or not. Every time I have tested this over the last year, my perceptions have been confirmed by the camera if the camera was set to sunny white balance. Then if I turn on a bright red flashlight, none of the colors look right. The dark night sky then looks black and as color perception first re-appears, the gray is a gray-blue. Slowly the blue goes away as the effects of the bright red light wears off, and the warm grays return.
Try this too at night in your house. Turn off all the lights so it is pretty dark and hard to see colors. Perhaps with window curtains open if there are no street lights outside. Dark adapt for at least 30 minutes. What colors can you see in the house? If you have a McBeth color chart, get it out. Now turn on a bright red flashlight and read a newspaper for a couple of minutes. Now what colors can you see? Hint: they will be really messed up.
So my conclusion is this fad is brought on by internet "teaching" by some who have warped their night vision with red flashlights.
It is a modern fad. After all, photographers never used tungsten film for night photography in the film era.
Roger
Posted by Roger Clark on August 26, 2014 at 6:20 PM
Hi Crystal,
One more thing on your image. The reflection in the lake does not match the sky and is much brighter than the sky (reflections are usually darker, or at most similar in brightness. So I would darken the lakes. Also, in a larger image (e.g. print), the star trails in the lakes may appear odd. You could clone out part of the trails, making them into points.
Roger
Posted by Grant Collier on August 26, 2014 at 6:39 PM
Hi Crystal,
This looks very good to me. There are some bright areas just above and below the horizon that you may want to tone down to make the blend look more natural. I do agree that the blues don't seem over-the-top. If you prefer a bit of blue to the sky like it is here, I recommend that you don't do this with a white balance adjustment. You can keep the white balance around 3800-4000K for a more neutral sky. Then, in Photoshop you can add more blue to the sky using a Color Balance adjustment layer with one of Tony Kuyper's Expanded Darks luminosity masks. This mask will prevent the adjustment from affecting bright areas of the sky, and you will be able to retain more natural colors in the stars and the Milky Way.
As for the noise in the sky, that is hard to eliminate. I use Topaz DeNoise, as I've found it does a very good job with reducing noise. I often take large stitched images of the sky to minimize noise. This can allow you to print the images much bigger. Another option is to use an equatorial mount, though this is more complicated.
WARNING: The rest of this is more in response to Jack & Roger and may put you to sleep if you're not a night photography addict...
For colors in the sky, there is some artistic license, as Jack says. Our eyes have color constancy, which means that we have subjective color perception depending on the color of the light illuminating a scene. Our eyes will try to make whites appear white regardless of whether it is lit by yellow or blue light. As a result, the same wavelength of light can appear very different to us depending on the color temperature of the entire scene. As a result, there are no “correct” colors, especially at night when we can’t see most colors. At night, the color temperature tends to be warmer, so our eyes compensate by making colors appear cooler. That is if the colors are bright enough to be seen. You see mostly in black & white at night, since our rods which we use for night vision don't see colors. The rods do perceive blue wavelengths of light much more strongly than reds, though, so blues will appear brighter to us at night This known as the Purkinje effect and artists have long taken it into account when designing some of their works.
I've found that 3800K is a good white balance to use to make whites, like snow, appear white under no moon and outside of twilight hours. So I’ll often use this white balance. If you want to take the Purkinje effect into account more, you could use a lower temp, but I stopped doing this a few years ago. I do find that it makes everything too blue and you do lose some nice colors. Some of the older shots on my web site were processed this way, though others were taken under astronomical twilight.
If, on the other hand, you used daylight white balance in this situation, snow would appear yellow, which is not how our eyes would ever perceive snow. Also, when shooting the northern lights, a daylight white balance makes the lights appear much more yellow than I saw with my eyes. Here's an example of a north lights shot with daylight white balance: http://www.gcollier.com/NLights.jpg. The lights as I saw them (and I've never used a red light) were a very bright neon green, but in this shot they are more yellow than they appeared to me. Yes, if I'd used a white balance below 3000K, they would have looked too bluish-green, so I don't disagree with Roger there.
Under a moon or during twilight, a white balance closer to daylight (around 4600-5100K) tends to work better. The sky will appear blue in these conditions.
I do tend to disagree on using a lower ISO. I don’t have the knowledge of the inner-workings of cameras that Roger has, but I I’ve done a lot of comparison shots at different ISOs over the years and have never seen the stars lose color at ISO 6400, unless the stars are actually being overexposed. You can check for this on your histogram in Lightroom. Under no moon at f2.8 I’ve never overexposed the stars (you can bring the colors out more in post-processing or you could overexpose them then but that's a different story). Here’s a test shot I did at different ISOs with my Nikon D800e at f2.8. http://www.gcollier.com/ISOtest.jpg
Every shot is underexposed by 1 stop compared to the one after it. To compensate for this in post-processing, I increased the exposure of each image in Lightroom by 1 stop compared to the one after it (or decreased by a stop for the higher ISOs) in order to make an equal comparison.
There’s not a big difference between ISO 1600 and 6400. But the hot pixels are more prominent at 1600 and I see a little bit of red discoloration at 1600 from dark noise. This was taken at twilight, so there’s not much color in the stars. I did take a test shot after twilight with much more color in the stars and saw no difference in color between 1600 and 6400. I even put the images on the same layer and rapidly switched back and forth between each image and couldn’t see any color difference. If I can find that comparison shot, I can post it later. So really, the difference beyond 1600 is minimal, but since I see less noise at 6400, I use that as long as I’m not overexposing anything. If I’m shooting meteors, it’s easier to overexpose them, so I shoot them at 1600 to try and keep any color in them.
Posted by Stan Rose on August 27, 2014 at 12:31 PM
I just read David Kingham's ebook on this subject--another good resource!
Looks good to me, Crystal--sharpen the fg a bit and it will look better.
Posted by Doug Beezley on August 27, 2014 at 8:43 PM
Crystal,
This is a beautiful, inspiring image and any critique from me would just be picking nits. I really like the composition and your exposure and processing bring out the best of the night. Your foreground exposure does its job - it connects the mysteries of the universe with what we know - the Earth.
Roger, your observations are interesting but I think most would view this and other images from an artistic point of view and not necessarily a scientific one. I am, however, puzzled by your statement that "there really are no blue stars". I just read an article that states all Class B stars are blue.
Doug Beezley
Snellville, GA
Posted by Roger Clark on August 27, 2014 at 10:19 PM
Hi Grant,
I agree with some of what you say, but you have some misconceptions. I have researched the eye for decades and have written a book on how the eye operates and detects objects in the night sky: "Visual Astronomy of the Deep Sky."
This is a good discussion because it is showing to me the thinking of other photographers. I pulled up my raw file for the image in my processing night photos article and tried your 3800 K method and found something fundamental which I'll go into below. This discussion is giving me ideas for my next article in the series: photographing the night sky and human color perception.
Grant Collier:
"For colors in the sky, there is some artistic license, as Jack says. Our eyes have color constancy, which means that we have subjective color perception depending on the color of the light illuminating a scene. Our eyes will try to make whites appear white regardless of whether it is lit by yellow or blue light. As a result, the same wavelength of light can appear very different to us depending on the color temperature of the entire scene. As a result, there are no “correct” colors, especially at night when we can’t see most colors. At night, the color temperature tends to be warmer, so our eyes compensate by making colors appear cooler. That is if the colors are bright enough to be seen. You see mostly in black & white at night, since our rods which we use for night vision don't see colors. The rods do perceive blue wavelengths of light much more strongly than reds, though, so blues will appear brighter to us at night This known as the Purkinje effect and artists have long taken it into account when designing some of their works."
Yes, the eye has a limited "auto" white balance. This could be a long discussion, so I'll get to the main point, which is the eye+brain wants a reference point. In the case of a colored light source illuminating a scene (e.g. red sunset) and you hold up a piece of what paper, we pick that out as white but do not see it as white, we can tell it has a red tint, but the auto white balance of the human system will make it look more white than a camera set on almost any white balance (sunny, cloudy, shade). The auto white balance on the camera, however will make the paper look whiter than we see with our eyes.
But the night sky is not illuminated by a single light source, rather it is billions of light source of all colors. Among those light sources are many bright stars. Those stars, especially the high temperature stars (like our sun and hotter) give us the reference points for the human system: they are the white reference points. If all stars were red, our system would get skewed, but because stars do cover the visible spectrum, and are bright, we have plenty of white reference points at night under a dark starry sky.
You are misinterpreting the Purkinje effect. The Perkinje effect is the shift in peak wavelength from the color sensing cones to the no color rods. The rods peak at shorter wavelengths and are more sensitive to green and blue light but see no color. So when light levels fall, color goes away even though blue sensitivity is higher. We simply do not see blue better when light levels fall, in fact it is the opposite.
See:
Shin et al., 2004a, Change of Color Appearance in Photopic, Mesopic and Scotopic Vision, Optical Review Vol. 11, pp. 265–271.
http://vision-lab.tp.chiba-u.jp/~yaguchi/MyPapers/2004(OR)_Change%20of%20color%20appearance%20in%20photopic,%20mesopic%20and%20scotopic%20vision.pdf
and
Shin et al., 2004b, A Color Appearance Model Applicable in Mesopic Vision, Optical Review Vol. 11, pp. 272–278.
http://vision-lab.tp.chiba-u.jp/~yaguchi/MyPapers/2004(OR)_A%20color%20appearance%20model%20applicable%20in%20mesopic%20vision.pdf
Basically, on a dark starry night we are working in the mesopic range of light, where both rods and cones are working. Brighter stars are at the higher end of the mesopic range, perhaps into the full color photopic range. But the Milky Way and airglow (unless very very bright) is in the mesopic range where color is draining away as the light levels drop. As shown by Shin et al., the blue and yellow sensitivity drops away before red-green sensitivity. I have confirmed this myself using a MacBeth color chart reducing light in a darkened room to the levels where I need the same exposure I use on my night sky images (30 seconds, f/1.4 ISO 1600). I also did a bunch of experiments late last night. At 10 seconds f/1.4, ISO 1600 (histogram peaks at about 1/3 on the scale) I could easily discern many colors, including yellow, green, red and cyan patches, but they appeared as pastels, not the saturated colors on the chart we see in normal room light. At 30 seconds, f/1.4, ISO 1600, after 20+ minutes of dark adaption, I could discern color in the green, red, and yellow, but the cyan was difficult to tell from yellow, and the no blues could be discerned. My views were similar to Shin et al, 2004b, their Figure 7.
So the bottom line, published research, as well as my own verification of that research, shows that as the light levels fall, blue sensitivity falls before red and green. Besides, if the Perkinje effect was a shift in color, the scotopic peak wavelength is more green than blue, so the sky would appear green not blue. But we know the effect of decreasing brightness dose not produce that color--rather it produces gray.
Further, there seems to be a lot of confusion on what the sensitivity is to light by the various cones. Some plots show the blue much higher than red and green, but those are not spectral sensitivity plots. See figure 5 here for the correct sensitivities:
Stockman, A. (2004). Colorimetry. In: T.G. Brown, K. Creath, H. Kogelnik, M.A. Kriss, J. Schmit, & M.J. Weber (Eds.), The Optics
Encyclopedia: Basic Foundations and Practical Applications, 1 (pp. 207-226). Berlin: Wiley-VCH.
http://www.cvrl.org/people/Stockman/pubs/2003%20Colorimetry%20S.pdf
Grant Collier:
"I've found that 3800K is a good white balance to use to make whites, like snow, appear white under no moon and outside of twilight hours. So I’ll often use this white balance. If you want to take the Purkinje effect into account more, you could use a lower temp, but I stopped doing this a few years ago. I do find that it makes everything too blue and you do lose some nice colors. Some of the older shots on my web site were processed this way, though others were taken under astronomical twilight."
If one wants to take the Purkijne effect into account, you would desaturate the scene, NOT use a lower temperature color balance. Color perception is being lost as the Purkinje effect comes into play. Here is my problem with the low color temperatures, both from a physics of the real world and color perception view. Changing color temperature is a color balance change, which in the digital file is a multiplication. But this changes the color of everything in the image, including bright stars, intermediate stars and faint stars. We can perceive the color of many stars with our unaided eyes, and with binoculars or a telescope, many more, as we can gather enough light to see their color.
But light pollution, scattering from aerosols (haze), and airglow are ADDED light to the scene. Thus, if one wants to compensate for the effects of any of these processes in the night sky, the proper method is to SUBTRACT them from the scene. You can see the effects of the different processes in the raw converter: changing the color temperature moves the RGB histograms relative to each other. When I use 3800 K for the scene in my processing page, the bright side of the histogram has blue higher, and bright red stars like Antares lose most of their red/orange color, and definitely not what I can easily see in the night sky with my unaided eyes.
So to keep bright star colors they way we perceive them and to subtract light pollution, scattering from haze, etc, one should SUBTRACT light from the image. In my processing thread,
http://www.clarkvision.com/articles/night.photography.image.processing/
I move the left side of the histogram in a method that does not clip the low end, but leaves intact the relative RGB relationships on the right side of the histogram, thus preserving star color closer to they way we perceive them.
If one wants to keep the colors of stars the way we perceive them, choose a white balance that does that (sunny, daylight) does that, and move the left side of the histogram to remove other effects. That is the way to make the sky blue (move only the left side of the histograms so that the red and green channels are lower than the blue) and keep bright stars accurate to what we perceive. Of course with artistic license, one can do anything one wants, including purple skies with purple stars.
Grant Collier:
"If, on the other hand, you used daylight white balance in this situation, snow would appear yellow, which is not how our eyes would ever perceive snow. Also, when shooting the northern lights, a daylight white balance makes the lights appear much more yellow than I saw with my eyes. Here's an example of a north lights shot with daylight white balance: http://www.gcollier.com/NLights.jpg. The lights as I saw them (and I've never used a red light) were a very bright neon green, but in this shot they are more yellow than they appeared to me. Yes, if I'd used a white balance below 3000K, they would have looked too bluish-green, so I don't disagree with Roger there."
The land is a different problem than the sky with stars, and the color on the land varies depending on the dominant color of the airglow. But I would ask, do you change the color of sunsets, including alpenglow so that the snow is always white? Under green auroral arcs in the arctic, do you change the color balance to make the snow white?
Grant Collier:
"Under a moon or during twilight, a white balance closer to daylight (around 4600-5100K) tends to work better. The sky will appear blue in these conditions."
"I do tend to disagree on using a lower ISO. I don’t have the knowledge of the inner-workings of cameras that Roger has, but I I’ve done a lot of comparison shots at different ISOs over the years and have never seen the stars lose color at ISO 6400, unless the stars are actually being overexposed. You can check for this on your histogram in Lightroom. Under no moon at f2.8 I’ve never overexposed the stars (you can bring the colors out more in post-processing or you could overexpose them then but that's a different story). Here’s a test shot I did at different ISOs with my Nikon D800e at f2.8.
http://www.gcollier.com/ISOtest.jpg"
I brought your image into photoshop and checked values. Most of the bright stars are saturated, even at ISO 100, and all images at all ISOs.
Grant Collier:
"There’s not a big difference between ISO 1600 and 6400. But the hot pixels are more prominent at 1600 and I see a little bit of red discoloration at 1600 from dark noise. This was taken at twilight, so there’s not much color in the stars. I did take a test shot after twilight with much more color in the stars and saw no difference in color between 1600 and 6400. I even put the images on the same layer and rapidly switched back and forth between each image and couldn’t see any color difference. If I can find that comparison shot, I can post it later. So really, the difference beyond 1600 is minimal, but since I see less noise at 6400, I use that as long as I’m not overexposing anything. If I’m shooting meteors, it’s easier to overexpose them, so I shoot them at 1600 to try and keep any color in them."
The color of bright stars should not change substantially in deep twilight if they are significantly brighter than the background, as you show in your images. You are not seeing star color because of image saturation. At the low ISO's this is due to your processing methods. At the high ISO is is likely both in camera saturation and post processing.
I did one other experiment last night. After I completed the color tests with no lights, I turned on a red led flashlight behind me shining on the wall behind me, so I did not look at the light, only the color target. The exposure on the target was about 0.1 second, f/1.4, ISO 1600. After about 20 seconds exposure to the red light, I turned it off and waited for dark adaption to return. After couple of minutes, I could adequately see all the squares on the color chart, including seeing the difference between the black square and the dark gray one next to it, as a slight change in intensity. I could discern no correct colors. However, for about 15 more minutes, continuing to dark adapt, I never could see correct colors and at first some of the lighter patches would flicker pastel pink and blue/cyan, with the squares alternating as I moved my head around. After 18 minutes, my color vision had not returned. Most of the pink/blue effect had gone away (but not entirely). I then stopped the observation. I was very surprised to see how long it took to recover color perception after exposure to red light.
There actually is evidence for this opponent color flickering. See:
Stockman et al, 2005, Spectrally opponent inputs to the human luminance pathway: slow +M and -L cone inputs revealed by
intense long-wavelength adaptation, J. Physiol v. 566.1, pp 61–76.
My conclusion is red lights really do mess up night color vision, and there is evidence they induce a blue response where none exists. Without such lights (and no bright light of any kind) we really can see colors in the night sky and they are consistent with what the camera records on sunny white balance.
Finally, for an an awe inspiring body of work on the human visual system, seethe work of Andrew Stockman:
http://www.cvrl.org/people/Stockman/Stockmanpub.htm
Roger
Posted by Roger Clark on August 27, 2014 at 10:27 PM
Doug:
"Roger, your observations are interesting but I think most would view this and other images from an artistic point of view and not necessarily a scientific one. I am, however, puzzled by your statement that "there really are no blue stars". I just read an article that states all Class B stars are blue."
Yes, this is confusing. Astronomers talk about blue stars (spectral class O, B) but only in the sense that the black body peak is in the blue. But such stars are "white hot" in that they emit so much at all visible colors that they appear white, or perhaps bluish white. You can see this in a plot of black body temperature on the chromaticity diagram, example here: http://en.wikipedia.org/wiki/Color_temperature
Note the 10,000K line is just a tiny bit on the blue side of white.
Even following the color temperature line to infinity does not bring one to the blue stars shown in many night photographs these days, including the image that started this thread.
Roger
Posted by Roger Clark on August 28, 2014 at 12:19 AM
I added to my processing web page how to create a blue night sky and keep star colors accurate in the way we perceive them, thus a combination natural and artistic license. As I described above color balance is not the physically correct way to process the images.
See the section called: "But I Really Want Blue Skies" and Figure 12a, 12b. at:
http://www.clarkvision.com/articles/night.photography.image.processing/
Roger
Posted by Grant Collier on August 28, 2014 at 2:04 AM
Roger, I wasn't expecting to change your mind. Just putting out a different viewpoint for others since you state things in a very black and white manner. I wasn't misinterpreting the Purkinje effect, I just didn't go into too much detail. At night, what you see is going to be much different than what the camera sees, so that's where some artistic license comes in. I was basically saying you could try to portray the world as our eyes might see it IF our rods were color sensitive (which they're not). I'd said this in more specificity in a reply to your post a few months ago. Alternatively, you could portray it like our eyes see it if our cones were ultra-light sensitive (which, again, they're not), in which case it would be more like we see colors during the day. We can see some colors at night, but it is quite limited. Neither your shots nor my shots represent what we actually see at night. Both have far more colors and detail than we can actually see. There's different ways of interpreting what the camera "sees" and converting that into a photograph. There is no fixed reality. Everyone sees the world differently and the camera sees the world much differently at night than we do (since it mimics how our cones see, not our rods). I think everyone should try to develop their own style, and it would be very boring if there was only one "right" way to do things.
As for the ISO test, again I stated that I've done a lot of tests over many years, not just this one. Others showed a lot more colors in the stars...it's not how I processed them, but the conditions I took them in. This particular shot was not in deep twilight but in earlier twilight with a moon. I have never seen more color in stars in ISO 1600 shots than in ISO 6400 shots. What you say sounds nice in theory, but I've yet to do a single test shot that coincides with what you say (assuming you reduce the exposure of the ISO 6400 shot by two stops to give an equal comparison). I'll take experimental results over theory any day. If you’ve actually done tests and see a difference in yours, maybe it's a result of using a different camera. But since I can see no difference, except slightly reduced noise at 6400, I’m certain no one else will see a difference in my shots either.
With some of your post-processing statements, you again say the “proper” method as if that is the only correct method. I was stating a method you could use if you prefer to artistically enhance the photo to make the skies bluer and keep some colors in the stars. Since I was stating a way of artistically manipulating the photos, there obviously is no one “proper” method. The world is not that black and white. I’m assuming you’ve never used luminosity masks, since you dismiss them so quickly. But they are a very powerful tool and can let you go beyond simple Curves adjustments to work specifically on certain tones of your image without affecting or blowing out other things like stars.
As for red lights, again I've never used a red light. The northern lights still appear much greener to me than they do with daylight white balance. People can see colors differently, so it's conceivable you'd see them as more yellow. The colors you say you see of stars and other objects in the night sky are very dim and it would be impossible to exactly match that to a star in a photo. There's a lot of subjectivity there. I’m not saying 3800K is the “right” way to do things. It’s how I’ve come to prefer shooting, partly because I’m also taking into account the landscape.
You say "The land is a different problem than the sky with stars." Not really, it’s all interrelated. If we could see perfectly well at night, our eyes would take the whole scene into view to determine colors and would likely use snow as a baseline for white, thus changing our perception of all the colors, not just the land. The stars would likely then appear a little less orange. If you’re only shooting the stars, I’d be more okay with the daylight white balance. But if I include the land, I take this into account and don’t want extreme color casts over the land that I never see during the day.
You stated before that our eyes “auto-white balance” can be limited, but it still exists. Using daylight white balance seems rather arbitrary. Would you use the daylight white balance indoors under incandescent lighting because our eye’s auto white balance effect is limited? Moonlight has a measured temperature of about 4150K. Would you again use daylight for this? If our eyes auto white balance is limited, perhaps you'd do something in between the two to account for this. Check the CMYK values of snow in some of your night shots. Your snow is very magenta in one shot and very yellow in most others. The C values are very low. Even with limited auto-white balance of our eyes, the fact that color constancy is a real phenomena dictates that we would never snow that off-color if we could see colors better at night. If you want to artistically interpret the colors that way, that's fine. I have no problem with that. Like you said, with artistic license you can make "purple skies with purple stars" or you can make magenta snow. And, your sunset analogy isn't accurate, since the clouds are lit directly by sunlight and the land is in shade, so there are two different lighting temperatures. Our eyes use whites in the shade to help determine color and see the reds in the sky as redder. The green aurora is a more extreme example and our eye's color constancy clearly can not completely counteract that. But, yes, I do still subdue the greens in the snow some to reflect color constancy. But the aurora is much more extreme than the color temperature of the starlit sky and keeping it exactly at Daylight when including the land in your shots w/out at all accounting for color constancy will, in my opinion, give you far too much of a yellow color cast.
Okay, I promise not to post anymore on this subject. Roger can have the last word if he likes and I’ll leave it be. I know Roger and I won’t ever agree completely and don’t want anything to become too heated.
Posted by Roger Clark on August 28, 2014 at 11:44 AM
Ok grant, I'll only make one comment, and that is regarding layer masks. Yes, I understand layer masks but what I show on my processing page with curves is exactly the same thing without the added complexity of luminosity masks, and with much more flexibility. For example, see Tony Kuyper's excellent tutorial on how to use luminosity masks: http://goodlight.us/writing/luminositymasks/luminositymasks-6.html
Just look at the complexity he had to go through to do mid-tones, highlights and shadows. Well, one does the same thing with curves and limiting where on the curve one deviates from the straight line. If I want to change the brightness level of the 20 to 25 range (on a 0 to 255 scale), simply select that part of the curves tool and change it while pinning the rest of the line to be straight. Simple, quick, no fuss, and much more specific to a problem than luminosity masks.
Thank you all for your insightful comments! I will try a test print and will work on reducing the brightness of the lakes. I have to say that this is perhaps the 5th time I have photographed the night sky with any sort of success, so Roger's discussion is certainly over my head. I'll reference it in the future though I'm sure!
Discussion
14 repliesFirst of all, my blend isn't the cleanest - so if you have suggestions I'd love to hear them. Second, the sky shot was taken at ISO 6400 and is thus quite noisy. How can I avoid producing an overly grainy photo? I've done plenty of noise reduction in Lightroom but still see quite a bit of grain. I will not be printing very large - 12x18 at the most with the remote possibility of 16x24.
Image stats:
Nikon D5100
Tokina 11-16mm
11 mm
foreground - ISO 3200 approximately 8 minutes
sky - ISO 6400 20 seconds
f/2.8
Thank you all SO much!
Beautiful image and at least not too extreme on the color balance (though popular these days to have blue stars on a blue sky--note there really are no blue stars and the Milky Way is not blue). You do have some nice reddish brown in the Milky way. The star Anatares is reddish-orange and does appear orange to the eye, but is white/blueish-white in your image. Similar with other stars: colors can be seen in brighter stars, more with binoculars.
Lots to discuss here. Not sure what you did in post processing, but ISO 6400 didn't help--most of the stars are saturated. ISO does not change sensitivity. So changing ISO will not change the noise, only exposure time and lens aperture will help that. I have a new article on post processing nightscapes to maintain star color:
http://www.clarkvision.com/articles/night.photography.image.processing/
For two exposures, land and sky, I keep them on separate layers. That can help avoid the bright rim around the mountain tops when trying to work
the image of the sky versus land.
Roger
Roger
Roger - Your articles just blew my mind! Very enlightening (no pun intended) information and explanations of ISO. It's amazing, and humbling, that after over a decade of digital photography I only now realize that I misunderstood ISO. Quite enlightening to learn that boosting the ISO past a certain point is not doing anything helpful to the image, and that it is better done in post processing after that point. I spent some time testing this in Lightroom with various night shots at same settings but different ISOs, and sure enough your explanations hold true. So thank you for that. Your website has a wealth of useful information.
One thing I would say - and I'm sure you've heard this a thousand times - is that the bluer night sky photos are perhaps closer to our perceptual reality. I understand and respect your statements about true sky/star colors, but also no human eye can see the galaxies and stars so clear as a camera can so we're already in "unreal" territory at this point. I would argue that the (slightly) bluer treatment is or can be closer to our subjective/perceptual reality of star gazing than a redder (though scientifically truer) rendition of the night sky.
The same idea would apply to moonlight photos (though I realize this is a different topic and more extreme example). When properly exposed, a moonlit scene might look just like daylight. But by processing the photo much bluer (for a "whiter" light rendition), the photo will more accurately convey the experience of how we see in the moonlight.
Your article here ( http://www.clarkvision.com/articles/color.of.the.night.sky/ ) is a great read and explains this issue thoroughly. Of particular interest related to my own statements would be your examples in Figures 10a and 10b. 10b is clearly the most beautiful and as you say accurate color rendering; however, were this my photo, I would be very much tempted to shift the background brown color (in the blacks) ever so slightly towards blue (not nearly as dramatically as 10a) which in my own opinion would appear more realistic to what my eyes would see (even if it's not).
It's a very interesting topic to me (thus my lengthy comments here!), and a topic you've tackled more thoroughly than any other I've run across.
You say "One thing I would say - and I'm sure you've heard this a thousand times - is that the bluer night sky photos are perhaps closer to our perceptual reality."
Well, I'm not sure a thousand times, but yes i have heard it a few times. But it is a recent, last few years fad, brought on in the digital era. And I would say the perception is highly colored (pun intended as you will see).
Let me go back to your lunar alpenglow image: beautiful!!!! (perhaps just a little too blue ;-) When the Moon is up in the sky and more than just a thin crescent, we do see Rayleigh scattering and a blue night sky.
But with no Moon and away from city lights, if one wanted to color their images the way they are perceived, it should be gray, not blue. I have studied the night sky for over 50 years. In the last year because of these issues, I've studied what colors can I see in the dark night sky. My conclusions so far as that people are seeing blue probably because of the bright red lights they are using. A strong color will give a strong opposite color response. There is another myth that red light won't hurt night vision. Yes they do, and it warps color perception too.
Try this experiment next time you are out under a dark, moonless night sky. Use no lights of any kind for 30 minutes. The sky will be mostly gray unless there is really strong airglow or aurora. Now, is the gray a warm gray, or a cool gray? When I do this, I become very well dark adapted with no light for 30 minutes, an hour better, I can see the color tint of the gray. Many stars show color, from blueish-white to yellow, orange and red. The core of the Milky Way in Sagittarius appears brownish gray. I can see banded airglow and can pick out the bands that are redder and those that are greener. If there is some auroral glow in the north, even when faint, I can pick out the color. A quick (10 to 30 second) exposure on a digital camera shows me if I am right or not. Every time I have tested this over the last year, my perceptions have been confirmed by the camera if the camera was set to sunny white balance. Then if I turn on a bright red flashlight, none of the colors look right. The dark night sky then looks black and as color perception first re-appears, the gray is a gray-blue. Slowly the blue goes away as the effects of the bright red light wears off, and the warm grays return.
Try this too at night in your house. Turn off all the lights so it is pretty dark and hard to see colors. Perhaps with window curtains open if there are no street lights outside. Dark adapt for at least 30 minutes. What colors can you see in the house? If you have a McBeth color chart, get it out. Now turn on a bright red flashlight and read a newspaper for a couple of minutes. Now what colors can you see? Hint: they will be really messed up.
So my conclusion is this fad is brought on by internet "teaching" by some who have warped their night vision with red flashlights.
It is a modern fad. After all, photographers never used tungsten film for night photography in the film era.
Roger
One more thing on your image. The reflection in the lake does not match the sky and is much brighter than the sky (reflections are usually darker, or at most similar in brightness. So I would darken the lakes. Also, in a larger image (e.g. print), the star trails in the lakes may appear odd. You could clone out part of the trails, making them into points.
Roger
This looks very good to me. There are some bright areas just above and below the horizon that you may want to tone down to make the blend look more natural. I do agree that the blues don't seem over-the-top. If you prefer a bit of blue to the sky like it is here, I recommend that you don't do this with a white balance adjustment. You can keep the white balance around 3800-4000K for a more neutral sky. Then, in Photoshop you can add more blue to the sky using a Color Balance adjustment layer with one of Tony Kuyper's Expanded Darks luminosity masks. This mask will prevent the adjustment from affecting bright areas of the sky, and you will be able to retain more natural colors in the stars and the Milky Way.
As for the noise in the sky, that is hard to eliminate. I use Topaz DeNoise, as I've found it does a very good job with reducing noise. I often take large stitched images of the sky to minimize noise. This can allow you to print the images much bigger. Another option is to use an equatorial mount, though this is more complicated.
WARNING: The rest of this is more in response to Jack & Roger and may put you to sleep if you're not a night photography addict...
For colors in the sky, there is some artistic license, as Jack says. Our eyes have color constancy, which means that we have subjective color perception depending on the color of the light illuminating a scene. Our eyes will try to make whites appear white regardless of whether it is lit by yellow or blue light. As a result, the same wavelength of light can appear very different to us depending on the color temperature of the entire scene. As a result, there are no “correct” colors, especially at night when we can’t see most colors. At night, the color temperature tends to be warmer, so our eyes compensate by making colors appear cooler. That is if the colors are bright enough to be seen. You see mostly in black & white at night, since our rods which we use for night vision don't see colors. The rods do perceive blue wavelengths of light much more strongly than reds, though, so blues will appear brighter to us at night This known as the Purkinje effect and artists have long taken it into account when designing some of their works.
I've found that 3800K is a good white balance to use to make whites, like snow, appear white under no moon and outside of twilight hours. So I’ll often use this white balance. If you want to take the Purkinje effect into account more, you could use a lower temp, but I stopped doing this a few years ago. I do find that it makes everything too blue and you do lose some nice colors. Some of the older shots on my web site were processed this way, though others were taken under astronomical twilight.
If, on the other hand, you used daylight white balance in this situation, snow would appear yellow, which is not how our eyes would ever perceive snow. Also, when shooting the northern lights, a daylight white balance makes the lights appear much more yellow than I saw with my eyes. Here's an example of a north lights shot with daylight white balance: http://www.gcollier.com/NLights.jpg. The lights as I saw them (and I've never used a red light) were a very bright neon green, but in this shot they are more yellow than they appeared to me. Yes, if I'd used a white balance below 3000K, they would have looked too bluish-green, so I don't disagree with Roger there.
Under a moon or during twilight, a white balance closer to daylight (around 4600-5100K) tends to work better. The sky will appear blue in these conditions.
I do tend to disagree on using a lower ISO. I don’t have the knowledge of the inner-workings of cameras that Roger has, but I I’ve done a lot of comparison shots at different ISOs over the years and have never seen the stars lose color at ISO 6400, unless the stars are actually being overexposed. You can check for this on your histogram in Lightroom. Under no moon at f2.8 I’ve never overexposed the stars (you can bring the colors out more in post-processing or you could overexpose them then but that's a different story). Here’s a test shot I did at different ISOs with my Nikon D800e at f2.8.
http://www.gcollier.com/ISOtest.jpg
Every shot is underexposed by 1 stop compared to the one after it. To compensate for this in post-processing, I increased the exposure of each image in Lightroom by 1 stop compared to the one after it (or decreased by a stop for the higher ISOs) in order to make an equal comparison.
There’s not a big difference between ISO 1600 and 6400. But the hot pixels are more prominent at 1600 and I see a little bit of red discoloration at 1600 from dark noise. This was taken at twilight, so there’s not much color in the stars. I did take a test shot after twilight with much more color in the stars and saw no difference in color between 1600 and 6400. I even put the images on the same layer and rapidly switched back and forth between each image and couldn’t see any color difference. If I can find that comparison shot, I can post it later. So really, the difference beyond 1600 is minimal, but since I see less noise at 6400, I use that as long as I’m not overexposing anything. If I’m shooting meteors, it’s easier to overexpose them, so I shoot them at 1600 to try and keep any color in them.
Looks good to me, Crystal--sharpen the fg a bit and it will look better.
This is a beautiful, inspiring image and any critique from me would just be picking nits. I really like the composition and your exposure and processing bring out the best of the night. Your foreground exposure does its job - it connects the mysteries of the universe with what we know - the Earth.
Roger, your observations are interesting but I think most would view this and other images from an artistic point of view and not necessarily a scientific one. I am, however, puzzled by your statement that "there really are no blue stars". I just read an article that states all Class B stars are blue.
Doug Beezley
Snellville, GA
I agree with some of what you say, but you have some misconceptions. I have researched the eye for decades and have written a book on how the eye operates and detects objects in the night sky: "Visual Astronomy of the Deep Sky."
This is a good discussion because it is showing to me the thinking of other photographers. I pulled up my raw file for the image in my processing night photos article and tried your 3800 K method and found something fundamental which I'll go into below. This discussion is giving me ideas for my next article in the series: photographing the night sky and human color perception.
Grant Collier:
"For colors in the sky, there is some artistic license, as Jack says. Our eyes have color constancy, which means that we have subjective color perception depending on the color of the light illuminating a scene. Our eyes will try to make whites appear white regardless of whether it is lit by yellow or blue light. As a result, the same wavelength of light can appear very different to us depending on the color temperature of the entire scene. As a result, there are no “correct” colors, especially at night when we can’t see most colors. At night, the color temperature tends to be warmer, so our eyes compensate by making colors appear cooler. That is if the colors are bright enough to be seen. You see mostly in black & white at night, since our rods which we use for night vision don't see colors. The rods do perceive blue wavelengths of light much more strongly than reds, though, so blues will appear brighter to us at night This known as the Purkinje effect and artists have long taken it into account when designing some of their works."
Yes, the eye has a limited "auto" white balance. This could be a long discussion, so I'll get to the main point, which is the eye+brain wants a reference point. In the case of a colored light source illuminating a scene (e.g. red sunset) and you hold up a piece of what paper, we pick that out as white but do not see it as white, we can tell it has a red tint, but the auto white balance of the human system will make it look more white than a camera set on almost any white balance (sunny, cloudy, shade). The auto white balance on the camera, however will make the paper look whiter than we see with our eyes.
But the night sky is not illuminated by a single light source, rather it is billions of light source of all colors. Among those light sources are many bright stars. Those stars, especially the high temperature stars (like our sun and hotter) give us the reference points for the human system: they are the white reference points. If all stars were red, our system would get skewed, but because stars do cover the visible spectrum, and are bright, we have plenty of white reference points at night under a dark starry sky.
You are misinterpreting the Purkinje effect. The Perkinje effect is the shift in peak wavelength from the color sensing cones to the no color rods. The rods peak at shorter wavelengths and are more sensitive to green and blue light but see no color. So when light levels fall, color goes away even though blue sensitivity is higher. We simply do not see blue better when light levels fall, in fact it is the opposite.
See:
Shin et al., 2004a, Change of Color Appearance in Photopic, Mesopic and Scotopic Vision, Optical Review Vol. 11, pp. 265–271.
http://vision-lab.tp.chiba-u.jp/~yaguchi/MyPapers/2004(OR)_Change%20of%20color%20appearance%20in%20photopic,%20mesopic%20and%20scotopic%20vision.pdf
and
Shin et al., 2004b, A Color Appearance Model Applicable in Mesopic Vision, Optical Review Vol. 11, pp. 272–278.
http://vision-lab.tp.chiba-u.jp/~yaguchi/MyPapers/2004(OR)_A%20color%20appearance%20model%20applicable%20in%20mesopic%20vision.pdf
Basically, on a dark starry night we are working in the mesopic range of light, where both rods and cones are working. Brighter stars are at the higher end of the mesopic range, perhaps into the full color photopic range. But the Milky Way and airglow (unless very very bright) is in the mesopic range where color is draining away as the light levels drop. As shown by Shin et al., the blue and yellow sensitivity drops away before red-green sensitivity. I have confirmed this myself using a MacBeth color chart reducing light in a darkened room to the levels where I need the same exposure I use on my night sky images (30 seconds, f/1.4 ISO 1600). I also did a bunch of experiments late last night. At 10 seconds f/1.4, ISO 1600 (histogram peaks at about 1/3 on the scale) I could easily discern many colors, including yellow, green, red and cyan patches, but they appeared as pastels, not the saturated colors on the chart we see in normal room light. At 30 seconds, f/1.4, ISO 1600, after 20+ minutes of dark adaption, I could discern color in the green, red, and yellow, but the cyan was difficult to tell from yellow, and the no blues could be discerned. My views were similar to Shin et al, 2004b, their Figure 7.
So the bottom line, published research, as well as my own verification of that research, shows that as the light levels fall, blue sensitivity falls before red and green. Besides, if the Perkinje effect was a shift in color, the scotopic peak wavelength is more green than blue, so the sky would appear green not blue. But we know the effect of decreasing brightness dose not produce that color--rather it produces gray.
Further, there seems to be a lot of confusion on what the sensitivity is to light by the various cones. Some plots show the blue much higher than red and green, but those are not spectral sensitivity plots. See figure 5 here for the correct sensitivities:
Stockman, A. (2004). Colorimetry. In: T.G. Brown, K. Creath, H. Kogelnik, M.A. Kriss, J. Schmit, & M.J. Weber (Eds.), The Optics
Encyclopedia: Basic Foundations and Practical Applications, 1 (pp. 207-226). Berlin: Wiley-VCH.
http://www.cvrl.org/people/Stockman/pubs/2003%20Colorimetry%20S.pdf
Grant Collier:
"I've found that 3800K is a good white balance to use to make whites, like snow, appear white under no moon and outside of twilight hours. So I’ll often use this white balance. If you want to take the Purkinje effect into account more, you could use a lower temp, but I stopped doing this a few years ago. I do find that it makes everything too blue and you do lose some nice colors. Some of the older shots on my web site were processed this way, though others were taken under astronomical twilight."
If one wants to take the Purkijne effect into account, you would desaturate the scene, NOT use a lower temperature color balance. Color perception is being lost as the Purkinje effect comes into play. Here is my problem with the low color temperatures, both from a physics of the real world and color perception view. Changing color temperature is a color balance change, which in the digital file is a multiplication. But this changes the color of everything in the image, including bright stars, intermediate stars and faint stars. We can perceive the color of many stars with our unaided eyes, and with binoculars or a telescope, many more, as we can gather enough light to see their color.
But light pollution, scattering from aerosols (haze), and airglow are ADDED light to the scene. Thus, if one wants to compensate for the effects of any of these processes in the night sky, the proper method is to SUBTRACT them from the scene. You can see the effects of the different processes in the raw converter: changing the color temperature moves the RGB histograms relative to each other. When I use 3800 K for the scene in my processing page, the bright side of the histogram has blue higher, and bright red stars like Antares lose most of their red/orange color, and definitely not what I can easily see in the night sky with my unaided eyes.
So to keep bright star colors they way we perceive them and to subtract light pollution, scattering from haze, etc, one should SUBTRACT light from the image. In my processing thread,
http://www.clarkvision.com/articles/night.photography.image.processing/
I move the left side of the histogram in a method that does not clip the low end, but leaves intact the relative RGB relationships on the right side of the histogram, thus preserving star color closer to they way we perceive them.
If one wants to keep the colors of stars the way we perceive them, choose a white balance that does that (sunny, daylight) does that, and move the left side of the histogram to remove other effects. That is the way to make the sky blue (move only the left side of the histograms so that the red and green channels are lower than the blue) and keep bright stars accurate to what we perceive. Of course with artistic license, one can do anything one wants, including purple skies with purple stars.
Grant Collier:
"If, on the other hand, you used daylight white balance in this situation, snow would appear yellow, which is not how our eyes would ever perceive snow. Also, when shooting the northern lights, a daylight white balance makes the lights appear much more yellow than I saw with my eyes. Here's an example of a north lights shot with daylight white balance: http://www.gcollier.com/NLights.jpg. The lights as I saw them (and I've never used a red light) were a very bright neon green, but in this shot they are more yellow than they appeared to me. Yes, if I'd used a white balance below 3000K, they would have looked too bluish-green, so I don't disagree with Roger there."
The land is a different problem than the sky with stars, and the color on the land varies depending on the dominant color of the airglow. But I would ask, do you change the color of sunsets, including alpenglow so that the snow is always white? Under green auroral arcs in the arctic, do you change the color balance to make the snow white?
Grant Collier:
"Under a moon or during twilight, a white balance closer to daylight (around 4600-5100K) tends to work better. The sky will appear blue in these conditions."
"I do tend to disagree on using a lower ISO. I don’t have the knowledge of the inner-workings of cameras that Roger has, but I I’ve done a lot of comparison shots at different ISOs over the years and have never seen the stars lose color at ISO 6400, unless the stars are actually being overexposed. You can check for this on your histogram in Lightroom. Under no moon at f2.8 I’ve never overexposed the stars (you can bring the colors out more in post-processing or you could overexpose them then but that's a different story). Here’s a test shot I did at different ISOs with my Nikon D800e at f2.8.
http://www.gcollier.com/ISOtest.jpg"
I brought your image into photoshop and checked values. Most of the bright stars are saturated, even at ISO 100, and all images at all ISOs.
Grant Collier:
"There’s not a big difference between ISO 1600 and 6400. But the hot pixels are more prominent at 1600 and I see a little bit of red discoloration at 1600 from dark noise. This was taken at twilight, so there’s not much color in the stars. I did take a test shot after twilight with much more color in the stars and saw no difference in color between 1600 and 6400. I even put the images on the same layer and rapidly switched back and forth between each image and couldn’t see any color difference. If I can find that comparison shot, I can post it later. So really, the difference beyond 1600 is minimal, but since I see less noise at 6400, I use that as long as I’m not overexposing anything. If I’m shooting meteors, it’s easier to overexpose them, so I shoot them at 1600 to try and keep any color in them."
The color of bright stars should not change substantially in deep twilight if they are significantly brighter than the background, as you show in your images. You are not seeing star color because of image saturation. At the low ISO's this is due to your processing methods. At the high ISO is is likely both in camera saturation and post processing.
I did one other experiment last night. After I completed the color tests with no lights, I turned on a red led flashlight behind me shining on the wall behind me, so I did not look at the light, only the color target. The exposure on the target was about 0.1 second, f/1.4, ISO 1600. After about 20 seconds exposure to the red light, I turned it off and waited for dark adaption to return. After couple of minutes, I could adequately see all the squares on the color chart, including seeing the difference between the black square and the dark gray one next to it, as a slight change in intensity. I could discern no correct colors. However, for about 15 more minutes, continuing to dark adapt, I never could see correct colors and at first some of the lighter patches would flicker pastel pink and blue/cyan, with the squares alternating as I moved my head around. After 18 minutes, my color vision had not returned. Most of the pink/blue effect had gone away (but not entirely). I then stopped the observation. I was very surprised to see how long it took to recover color perception after exposure to red light.
There actually is evidence for this opponent color flickering. See:
Stockman et al, 2005, Spectrally opponent inputs to the human luminance pathway: slow +M and -L cone inputs revealed by
intense long-wavelength adaptation, J. Physiol v. 566.1, pp 61–76.
My conclusion is red lights really do mess up night color vision, and there is evidence they induce a blue response where none exists. Without such lights (and no bright light of any kind) we really can see colors in the night sky and they are consistent with what the camera records on sunny white balance.
Finally, for an an awe inspiring body of work on the human visual system, seethe work of Andrew Stockman:
http://www.cvrl.org/people/Stockman/Stockmanpub.htm
Roger
"Roger, your observations are interesting but I think most would view this and other images from an artistic point of view and not necessarily a scientific one. I am, however, puzzled by your statement that "there really are no blue stars". I just read an article that states all Class B stars are blue."
Yes, this is confusing. Astronomers talk about blue stars (spectral class O, B) but only in the sense that the black body peak is in the blue. But such stars are "white hot" in that they emit so much at all visible colors that they appear white, or perhaps bluish white. You can see this in a plot of black body temperature on the chromaticity diagram, example here: http://en.wikipedia.org/wiki/Color_temperature
Note the 10,000K line is just a tiny bit on the blue side of white.
Even following the color temperature line to infinity does not bring one to the blue stars shown in many night photographs these days, including the image that started this thread.
Roger
See the section called: "But I Really Want Blue Skies" and Figure 12a, 12b. at:
http://www.clarkvision.com/articles/night.photography.image.processing/
Roger
As for the ISO test, again I stated that I've done a lot of tests over many years, not just this one. Others showed a lot more colors in the stars...it's not how I processed them, but the conditions I took them in. This particular shot was not in deep twilight but in earlier twilight with a moon. I have never seen more color in stars in ISO 1600 shots than in ISO 6400 shots. What you say sounds nice in theory, but I've yet to do a single test shot that coincides with what you say (assuming you reduce the exposure of the ISO 6400 shot by two stops to give an equal comparison). I'll take experimental results over theory any day. If you’ve actually done tests and see a difference in yours, maybe it's a result of using a different camera. But since I can see no difference, except slightly reduced noise at 6400, I’m certain no one else will see a difference in my shots either.
With some of your post-processing statements, you again say the “proper” method as if that is the only correct method. I was stating a method you could use if you prefer to artistically enhance the photo to make the skies bluer and keep some colors in the stars. Since I was stating a way of artistically manipulating the photos, there obviously is no one “proper” method. The world is not that black and white. I’m assuming you’ve never used luminosity masks, since you dismiss them so quickly. But they are a very powerful tool and can let you go beyond simple Curves adjustments to work specifically on certain tones of your image without affecting or blowing out other things like stars.
As for red lights, again I've never used a red light. The northern lights still appear much greener to me than they do with daylight white balance. People can see colors differently, so it's conceivable you'd see them as more yellow. The colors you say you see of stars and other objects in the night sky are very dim and it would be impossible to exactly match that to a star in a photo. There's a lot of subjectivity there. I’m not saying 3800K is the “right” way to do things. It’s how I’ve come to prefer shooting, partly because I’m also taking into account the landscape.
You say "The land is a different problem than the sky with stars." Not really, it’s all interrelated. If we could see perfectly well at night, our eyes would take the whole scene into view to determine colors and would likely use snow as a baseline for white, thus changing our perception of all the colors, not just the land. The stars would likely then appear a little less orange. If you’re only shooting the stars, I’d be more okay with the daylight white balance. But if I include the land, I take this into account and don’t want extreme color casts over the land that I never see during the day.
You stated before that our eyes “auto-white balance” can be limited, but it still exists. Using daylight white balance seems rather arbitrary. Would you use the daylight white balance indoors under incandescent lighting because our eye’s auto white balance effect is limited? Moonlight has a measured temperature of about 4150K. Would you again use daylight for this? If our eyes auto white balance is limited, perhaps you'd do something in between the two to account for this. Check the CMYK values of snow in some of your night shots. Your snow is very magenta in one shot and very yellow in most others. The C values are very low. Even with limited auto-white balance of our eyes, the fact that color constancy is a real phenomena dictates that we would never snow that off-color if we could see colors better at night. If you want to artistically interpret the colors that way, that's fine. I have no problem with that. Like you said, with artistic license you can make "purple skies with purple stars" or you can make magenta snow. And, your sunset analogy isn't accurate, since the clouds are lit directly by sunlight and the land is in shade, so there are two different lighting temperatures. Our eyes use whites in the shade to help determine color and see the reds in the sky as redder. The green aurora is a more extreme example and our eye's color constancy clearly can not completely counteract that. But, yes, I do still subdue the greens in the snow some to reflect color constancy. But the aurora is much more extreme than the color temperature of the starlit sky and keeping it exactly at Daylight when including the land in your shots w/out at all accounting for color constancy will, in my opinion, give you far too much of a yellow color cast.
Okay, I promise not to post anymore on this subject. Roger can have the last word if he likes and I’ll leave it be. I know Roger and I won’t ever agree completely and don’t want anything to become too heated.
Just look at the complexity he had to go through to do mid-tones, highlights and shadows. Well, one does the same thing with curves and limiting where on the curve one deviates from the straight line. If I want to change the brightness level of the 20 to 25 range (on a 0 to 255 scale), simply select that part of the curves tool and change it while pinning the rest of the line to be straight. Simple, quick, no fuss, and much more specific to a problem than luminosity masks.
Roger
Thanks!