Aurora - White Balance Experiment

To all,

Here is one of my aurora shots using different white balances. Even though the aurora may be more naturally colored when using a Daylight white balance, it results in an unrealistically colored background, at least for me. I could add a few hundred Kelvin to my current images, but I just can't get myself to sacrifice the background sky color. The sky is not a brownish red color. What are your thoughts? Which image do you prefer? Roger, I already know your preference; nevertheless, I would love to hear your thoughts.

Here is an explanation of how I derived my neutral background. I found an image where the aurora had disappeared or became nearly inactive. I picked an area where I thought would have the least likelihood of being affected by the aurora or other atmospheric phenomenon (highest point in the frame). I determined a temperature of 3450K with a tint of -20 produced the closest neutral sky background. I then used these values for all the images taken between the ending and the beginning of astronomical twilight.

The following are settings made in Lightroom.

As shot: Temp 3500; Tint +17


The camera determined the white balance.



Daylight balanced: Temp 5500; Tint +10


Setting the white balance to Daylight



Neutral background: Temp 3450K; Tint -20


Tweaked the white balance to get a neutral sky.



Daylight balanced: Temp 5500K; Tint +10


Sky background using a Daylight white balance.

Thanks for looking,

Wade

Hi Wade,

I didn't comment on your original post, so first of all I would like to say that I love this shot, as well as the others in your gallery! I often like these far away side-on views more than the typical looking-up views, since they offer a better glimps at the physics behind the Aurorae. (That said, some of the looking-up ones can be visually more impressive.)

On the topic of white balance. I must say I can enjoy all three versions, with a preference for both your handpicked WB and the dayligh WB. Of course Roger is right that the latter is the "correct" one, and I can enjoy it from an esthetic point of view, but your hand-balanced version might please more people as it shows a wide range of subtle colour gradients. Again, I can enjoy them both, and as long as we are aware which are the true colours, I see no problem in going down either route. One thing I would note is that the "true" colour can be discussed further as well, since our eyes will not see as much colour as our cameras can pick up: http://earthsky.org/earth/will-you-see-colors-in-an-aurora

Anyway, for someone that is partially colour blind, I'd say you do a pretty good job at colour-balancing your images ;-) One question: what camera profile did you use in Lightroom? (For my Canon 70D, I know from landscape pictures how much colours vary between the "Adobe Standard" profile and the different camera-specific ones - for the nightsky I usually start from "Camera Neutral" or in some cases "Camera Standard".)

Thanks for showing these,

regards,

--
Greg Van den Bleeken
www.pbase.com/gbleek
vimeo.com/vdbphotography
Take photographs *you* want to look at. Take photographs you want to *look* at. (Ed Leys)

Interesting, thanks for sharing. I have to admit to a preference for real colours, despite another looking 'prettier'.

I don't suppose there is a way to subtract light pollution from aurora, is there? I guess aurora is a form of LP.

Nate

t_wade said:

To all,

Here is one of my aurora shots using different white balances. Even though the aurora may be more naturally colored when using a Daylight white balance, it results in an unrealistically colored background, at least for me. I could add a few hundred Kelvin to my current images, but I just can't get myself to sacrifice the background sky color. The sky is not a brownish red color. What are your thoughts? Which image do you prefer? Roger, I already know your preference; nevertheless, I would love to hear your thoughts.

Here is an explanation of how I derived my neutral background. I found an image where the aurora had disappeared or became nearly inactive. I picked an area where I thought would have the least likelihood of being affected by the aurora or other atmospheric phenomenon (highest point in the frame). I determined a temperature of 3450K with a tint of -20 produced the closest neutral sky background. I then used these values for all the images taken between the ending and the beginning of astronomical twilight.

The following are settings made in Lightroom.

Daylight balanced: Temp 5500; Tint +10


Setting the white balance to Daylight

Click to expand...

Hi Wade,

The most accurate colors are the daylight white balance. The night sky is very rarely a neutral color due to ariglow and is often red or green or combinations of red and green (so yellows and oranges, all from oxygen emissions). In equatorial regions, red is more common, toward polar regions, green more common and mid-latitudes both. So when you tried to neutralize a sky background with white balance (a multiply), you are applying those multipliers to all the brightest parts of the scene too. That produces the varying color balance with scene intensity creating the blue/purple as the scene intensities drop. Instead, subtract from the sky background to neutralize it, then you will get accurate colors throughout the brightness range in the aurora.

Roger

Roger,

rnclark said:

The most accurate colors are the daylight white balance.

Click to expand...

There's definitely a certain appeal to the look. I'll likely choose a balance between the two images.

rnclark said:

So when you tried to neutralize a sky background with white balance (a multiply), you are applying those multipliers to all the brightest parts of the scene too. That produces the varying color balance with scene intensity creating the blue/purple as the scene intensities drop. Instead, subtract from the sky background to neutralize it, then you will get accurate colors throughout the brightness range in the aurora.

Click to expand...

What settings do you recommend using in Lightroom to achieve the same behavior as WB but in a subtractive oriented fashion? I would certainly like to experiment with the additive and subtraction settings, rather than multiplicative settings (i.e., WB).

I appreciate your expertise and will put your suggestions to good use.

Thanks,

Wade

P.S. I'm always torn between accuracy and "a pretty picture look". :-(

Greg,

Greg VdB said:

I didn't comment on your original post, so first of all I would like to say that I love this shot, as well as the others in your gallery!

Click to expand...

Thanks!

Greg VdB said:

I often like these far away side-on views more than the typical looking-up views, since they offer a better glimps at the physics behind the Aurorae.

Click to expand...

This is my view too; however, I haven't seen the aurora from below so I might be biased.

Greg VdB said:

On the topic of white balance. I must say I can enjoy all three versions, with a preference for both your handpicked WB and the dayligh WB. Of course Roger is right that the latter is the "correct" one, and I can enjoy it from an esthetic point of view, but your hand-balanced version might please more people as it shows a wide range of subtle colour gradients.

Click to expand...

I'm always torn between accuracy and "a pretty picture look". I'll likely settle for a good balance between the two.

Greg VdB said:

One thing I would note is that the "true" colour can be discussed further as well, since our eyes will not see as much colour as our cameras can pick up: http://earthsky.org/earth/will-you-see-colors-in-an-aurora

Click to expand...

I enjoyed the link you mentioned. It certainly puts a perspective on how poorly our eyes respond in low lit scenes.

Greg VdB said:

Anyway, for someone that is partially colour blind, I'd say you do a pretty good job at colour-balancing your images ;-)

Click to expand...

Thanks, it takes a lot of effort and time.

Greg VdB said:

One question: what camera profile did you use in Lightroom? (For my Canon 70D, I know from landscape pictures how much colours vary between the "Adobe Standard" profile and the different camera-specific ones - for the nightsky I usually start from "Camera Neutral" or in some cases "Camera Standard".)

Click to expand...

I'm not sure which one I use, probably the Standard since I haven't changed the defaults. I always shoot in RAW though.

Thanks for taking the time to respond to my query.

Wade

Nate,

BlackgumNate said:

Interesting, thanks for sharing. I have to admit to a preference for real colours, despite another looking 'prettier'.

Click to expand...

I'm torn between the two. I'll likely choose a compromise between the two.

BlackgumNate said:

I don't suppose there is a way to subtract light pollution from aurora, is there? I guess aurora is a form of LP.

Click to expand...

Probably not, as it would subtract partially from the aurora too.

I appreciate you taking the time to answer my query.

Wade

t_wade said:

Roger,

t_wade said:

The most accurate colors are the daylight white balance.

Click to expand...

There's definitely a certain appeal to the look. I'll likely choose a balance between the two images.

t_wade said:

So when you tried to neutralize a sky background with white balance (a multiply), you are applying those multipliers to all the brightest parts of the scene too. That produces the varying color balance with scene intensity creating the blue/purple as the scene intensities drop. Instead, subtract from the sky background to neutralize it, then you will get accurate colors throughout the brightness range in the aurora.

Click to expand...

What settings do you recommend using in Lightroom to achieve the same behavior as WB but in a subtractive oriented fashion? I would certainly like to experiment with the additive and subtraction settings, rather than multiplicative settings (i.e., WB).

Click to expand...

In any photo editor, use the curves tool and the lower left point, slide to the right by channel. The mount you slide to the right is the amount subtracted. Figure 1c and 1d here illustrates it: http://www.clarkvision.com/articles/astrophotography.image.processing2/

t_wade said:

I appreciate your expertise and will put your suggestions to good use.

Thanks,

Wade

P.S. I'm always torn between accuracy and "a pretty picture look". :-(

Click to expand...

Generally I would say we all want pretty images. In film days, color was so uncontrollable with long exposure reciprocity failure about anything could be the result. Now with digital more precision is easy. But in this digital age Pro non-astro photographers started teaching night photography and they pushed the idea that the night sky is blue, so much so that it is now ingrained as the truth (truthiness), but I and a few others are making inroads. It does look pretty (e.g. your purple aurora). But The night sky is so beautifully colored that I am finding that people are more impressed when they ask if that is natural color and I say yes. Plus even more impressed when I can relate the colors to the actual processes going on, e.g. red is oxygen emitting like a neon sign at 100 to 300 km altitude. If I show them a pretty image with unreal colors they express disappointment when told the colors are not real.

Reality can be fun.

Roger

rnclark said:

In any photo editor, use the curves tool and the lower left point, slide to the right by channel. The mount you slide to the right is the amount subtracted. Figure 1c and 1d here illustrates it: http://www.clarkvision.com/articles/astrophotography.image.processing2/

Click to expand...

I have been playing with curves in RawTherapee to subtract red and green to balance it with my memory. For sure it retains star colors much better. It also seems to work pretty well on other subjects instead of the white balance slider. But there seem to be several ways to do it.

(a) Move the darkest points in the RGB curves to the right to subtract light

This subtracts more light from the dark parts than from the bright parts, and the R,G, and B get different slopes (contrast).

(b) Move the darkest points in the RGB curves to the right to subtract light, but also move the brightest points down so that the straight lines are parallel.

This is real subtraction of a constant from all brightness values.

(c) Keep the darkest and brightest RGB points fixed, then pull the R and G down where the histogram has its peak.

This subtracts most light pollution from the main background brightness. It also introduces different slopes corresponding to contrast per color channel, in dark and brighter parts. This can show up as subtle red and green banding.

Also possible is:

(d) Hybrid method subtracting like (b), creating parallel straight lines over most of the range, but bending off towards the fixed darkest/brightest points.

This should keep odd gradients away but takes more points to do it, while it is difficult to keep parallel lines when tweaking. Any thoughts?

Secondly, after constructing these RGB curves, which method is the correct one for handling colors when tweaking the main contrast curve? See:


Actually is it possible in some software to create simple gradients per color channel to subtract light pollution? RawTherapee offers an awful lot of things, but not this, and I do not have Photoshop.

oscarvdvelde said:

rnclark said:

In any photo editor, use the curves tool and the lower left point, slide to the right by channel. The mount you slide to the right is the amount subtracted. Figure 1c and 1d here illustrates it: http://www.clarkvision.com/articles/astrophotography.image.processing2/

Click to expand...

I have been playing with curves in RawTherapee to subtract red and green to balance it with my memory. For sure it retains star colors much better. It also seems to work pretty well on other subjects instead of the white balance slider. But there seem to be several ways to do it.

(a) Move the darkest points in the RGB curves to the right to subtract light

This subtracts more light from the dark parts than from the bright parts, and the R,G, and B get different slopes (contrast).

(b) Move the darkest points in the RGB curves to the right to subtract light, but also move the brightest points down so that the straight lines are parallel.

This is real subtraction of a constant from all brightness values.

Click to expand...

This is correct IF the data are linear. Usually by the time one is using curves, a tone curve which compresses the upper end has been applied to the data. In that case it will not do the intended subtraction--it will be subtracting too much at the high end. In that case, keeping the high end fixed is better.

oscarvdvelde said:

(c) Keep the darkest and brightest RGB points fixed, then pull the R and G down where the histogram has its peak.

This subtracts most light pollution from the main background brightness. It also introduces different slopes corresponding to contrast per color channel, in dark and brighter parts. This can show up as subtle red and green banding.

Also possible is:

(d) Hybrid method subtracting like (b), creating parallel straight lines over most of the range, but bending off towards the fixed darkest/brightest points.

This should keep odd gradients away but takes more points to do it, while it is difficult to keep parallel lines when tweaking. Any thoughts?

Click to expand...

Per above, don't keep parallel lines on tone curve data.

oscarvdvelde said:

Secondly, after constructing these RGB curves, which method is the correct one for handling colors when tweaking the main contrast curve? See:

http://50.87.144.65/~rt/w/index.php?title=Exposure#Curve_Mode

Click to expand...

If you don't want color to change when you change contrast, then modification of the tone curve should be done on luminance (or the L channel in LAB mode).

oscarvdvelde said:

Actually is it possible in some software to create simple gradients per color channel to subtract light pollution? RawTherapee offers an awful lot of things, but not this, and I do not have Photoshop.

Click to expand...

It is easy in photoshop. Try the new 16-bit capable gimp.

Roger

t_wade said:

To all,

Here is one of my aurora shots using different white balances. Even though the aurora may be more naturally colored when using a Daylight white balance, it results in an unrealistically colored background, at least for me. I could add a few hundred Kelvin to my current images, but I just can't get myself to sacrifice the background sky color. The sky is not a brownish red color. What are your thoughts? Which image do you prefer? Roger, I already know your preference; nevertheless, I would love to hear your thoughts.

Here is an explanation of how I derived my neutral background. I found an image where the aurora had disappeared or became nearly inactive. I picked an area where I thought would have the least likelihood of being affected by the aurora or other atmospheric phenomenon (highest point in the frame). I determined a temperature of 3450K with a tint of -20 produced the closest neutral sky background. I then used these values for all the images taken between the ending and the beginning of astronomical twilight.

The following are settings made in Lightroom.

As shot: Temp 3500; Tint +17


The camera determined the white balance.

Daylight balanced: Temp 5500; Tint +10

Neutral background: Temp 3450K; Tint -20

Daylight balanced: Temp 5500K; Tint +10

Thanks for looking,

Wade

Click to expand...

I think the first [3500K] one looks best and more like the Aurora I saw on May 8th as well. I'm no expert, but I took some shots here in Vancouver B.C. [where Auroras rarely hit this far south..]

-Martin P





Aurora Borealis Van BC

To get a good Point of view what the "real" WB should be. Shoot at the opposite direction of the aura and use to White balance it correctly. Apply it to the other images and finetune.

/Stefan

Martin,

Lensmate said:

I think the first [3500K] one looks best and more like the Aurora I saw on May 8th as well. I'm no expert, but I took some shots here in Vancouver B.C. [where Auroras rarely hit this far south..]

Click to expand...

As Roger has noted, there's no way you can get these colors from an aurora. They emit light as specific wavelengths which our cameras and eyes are poor at "seeing". Cameras always get the white balance (WB) wrong. Even my usual technique of making the background neutral from "a near zenith point" doesn't work. This is all due to the fact aurora are emitting their on light.

I have settled on a WB with a temperature around 4400-4600K and a tint around 3-5 in Lightroom. I couldn't go with a Daylight balance for a more natural look since it made the sky background too red for my liking . As a person that is Red/Green colored blind, if I can see that the sky is too red it is really, really too red.

Keep in mind, there's certainly nothing wrong in using any WB during the post-processing phase; however, one has the option to strive for a more natural look.

Nice shot by the way.

Wade

Stefan,

Digitaliz said:

To get a good Point of view what the "real" WB should be. Shoot at the opposite direction of the aura and use to White balance it correctly. Apply it to the other images and finetune.

Click to expand...

This will not work if you want a natural look to your aurora images. Aurora emit light at specific wavelengths so determining your WB from a different source will always produce incorrect results. As Roger has noted, you need to choose a "Daylight" WB if you want a natural look to your aurora. I couldn't handle the red sky this produced so I opted to compromise with a custom WB using a temperature range of 4400-4600 and a tint of 3-5.

Wade

You don't obtain WB from a different source. It's like having a greycard with you. You will obtain the WB from the surrounding where you stand without having the auro in the image, making it easier to pick WB from a neutral grey, like i.e. from a stone.

Best is actually to shoot from the same location when there is no aura at all and use that WB to set when it is aurora borealis.

Setting it to daylight will only work if you are in a non light polluted area. It's the same as shooting the MW. The closer you are to light pollution, the more compensation will be needed to get rid of the city lights w´which do not light up the sky with daylight spotlights.

/Stefan

t_wade said:

Stefan,

t_wade said:

To get a good Point of view what the "real" WB should be. Shoot at the opposite direction of the aura and use to White balance it correctly. Apply it to the other images and finetune.

Click to expand...

This will not work if you want a natural look to your aurora images. Aurora emit light at specific wavelengths so determining your WB from a different source will always produce incorrect results. As Roger has noted, you need to choose a "Daylight" WB if you want a natural look to your aurora. I couldn't handle the red sky this produced so I opted to compromise with a custom WB using a temperature range of 4400-4600 and a tint of 3-5.

Wade

Click to expand...

Stefan,

Digitaliz said:

You don't obtain WB from a different source. It's like having a greycard with you. You will obtain the WB from the surrounding where you stand without having the auro in the image, making it easier to pick WB from a neutral grey, like i.e. from a stone.

Click to expand...

Yes, if you use a gray card, you will get a great WB for the landscape; however, the colors of the aurora will be completely "wrong" (i.e. not natural). As stated before, the aurora emit their own type of light at certain wavelengths which match closely to a "Daylight" WB. So in summary, during the aurora, you can either have a naturally colored aurora by choosing a "Daylight" WB or have a correct WB for the landscape/sky using your technique, but you can't have both. :-(

Hopefully, Roger will chime into the discussion and offer more detail.

Wade