Temperature and Tint

[Hening Bettermann]

The color temperature of daylight changes in a somewhat predictable way under certain conditions, such as time of the day and amount of clouds. Thus, a very rough color temperature scale can be made, like this from CambridgeInColour.com(artificial light sources omitted):
3000-4000 K    Sunrise/Sunset (clear sky)
5000-6500 K    Daylight with Clear Sky (sun overhead)
6500-8000 K    Moderately Overcast Sky
9000-10 000 K   Shade or Heavily Overcast Sky

These numbers refer to the red/yellow-blue/cyan axis.
Does the Tint (changes along the magenta-green axis) behave in a similar way?
So that it would be possible to correlate these two, say that when Temp is in the range 6500-8000, then tint is in the range +10..+20 ?

[Czornyj]

Of course not. Daylight is generally close to planckian locus curve, and there's only a slight tint change.

http://www.ugr.es/~colorimg/pdfs/ao_1999_5703.pdf

[Hening Bettermann]

Thank you, Czornyj, for your answer and the reference.

[Tim Lookingbill]

Is this information needed for editing images or calibrating a display?

[fetish]

it smells like it's related to adobe lightroom based image editing.

[Hening Bettermann]

>Is this information needed for editing images or calibrating a display?
>it smells like it's related to adobe lightroom based image editing.

My idea is to replace the As Shot white balance by my best memory of the kind of daylight at shooting time. My raw converter is Raw Developer, I don't have Lightroom.

Good light! - Hening.

[Tim Lookingbill]

>Is this information needed for editing images or calibrating a display?
>it smells like it's related to adobe lightroom based image editing.

My idea is to replace the As Shot white balance by my best memory of the kind of daylight at shooting time. My raw converter is Raw Developer, I don't have Lightroom.

Good light! - Hening.

Hope you have a good memory. Kelvin numbers are just a ballpark estimate in relation to the appearance of color temp cast as interpreted by the Raw converter.

I've used Raw Developer, (purchased it) but I'm not pleased with its rendering engine (I believe Lab) when it comes to trying to get "pleasing" colors only when editing. Usually RD's default rendering is a bit more realistic than Adobe Camera Raw's, my converter of choice.

I'ld suggest if the default RD preview looks neutral overall and balanced (no one color is over saturated compared to the rest) but is still just a bit dull, to increase saturation first before attempting to warm or cool with color temp adjusts.

[Hening Bettermann]

Hei tlooknbill

Not sure if you understand my intention. That is to come a little closer to a "theoretically correct" color. By that I mean the perceived color at shooting time. This is neither the color that would be achieved by using the grey card, nor any arbitrary "pleasing" color.

I don't find RD's default rendering dull - quite on the contrary, I find it (and it is admittedly) tailored to please - in my lexikon: over-saturated colors and over-steepened contrast. I have replaced it by my own home-made camera profiles. My images don't look good in the raw converter, and are not supposed to, because they are linear at this stage.

> Hope you have a good memory.
Alas, this is of course the Achilles heel of my intended "method". But even a rough scale like sun around noon - afternoon - near sunset - slightly overcast - heavy overcast may be more accurate than the camera's guess, as long as this guess is based on the image, not the light. Since I do landscape, the image itself often shows a bit of the sky to aid my memory. The EXIF data record the shooting time. Maybe I will in the future make field notes concerning the degree of cloudiness. So far I have used my FRESH memory, trying to establish WB the same evening or next day.

So what I plan to use is not my memory of the colors/the wb themselves. I hope to reconstruct the most accurate WB which is possible after the circumstances using a combination of time of the day and degree of cloudiness. 

Good light - Hening.

[madmanchan]

temp/tint numbers are only a rough guide.  The thing that really matters to the image is the spectrum of the illumination, which cannot be summarized by just a pair of numbers.  To make matters more complicated, lighting is rarely uniform over the entire scene.  So, a global temp/tint white balance is really a ballpark estimate to get you started.

[Tim Lookingbill]

So what I plan to use is not my memory of the colors/the wb themselves. I hope to reconstruct the most accurate WB which is possible after the circumstances using a combination of time of the day and degree of cloudiness.  

Would be curious to see how you go about doing that. Are you going to use measuring instruments at the time you shoot the landscape, record the number and then plug those numbers into Raw Developer?

Don't know if you're aware that the eyes perceive color cast not just from the filter effect of color temp but how real objects in a landscape reflect back their spectral make up that can contrast/compliment/amplify against this overall filter color cast effect. This usually shows up as saturation, but it's never quite determined if it's from the color cast or the spectral reflectance of the object itself in a scene. This is what makes relying on memory and at the same time trying to achieve "accuracy" not a productive part of image creation.

You are performing a magic trick in an attempt to mimic the spectral reflectance of each object in the scene lit by a natural light source such as the sun. Even at high noon direct sunlight there are objects in a scene that you would think would look neutral when on inspection with regards to accuracy may have quite a warm or cool hue, but it's hard to know whether it's from the color cast (which you should see none even at this time of day if accuracy is of concern) or the objects spectral reflectance characteristics.

This is why I suggested first increase saturation. There are some complicated optical effects going on in scene reconstruction in an artificial editing environment that aren't just about color temp that can defy numbers. Color Constancy to name one.

I've started out editing images of outdoor scenes that seem to be too blue and dull and think the dullness is from the overall blue cast. Because I didn't want to get into the back and forth fiddling of color temp/magenta/green sliders, I would increase saturation first and all of a sudden the blue cast went away.

What Kelvin number would that be because the Kelvin numbers didn't change, but my perception did?

[Czornyj]

temp/tint numbers are only a rough guide.  The thing that really matters to the image is the spectrum of the illumination, which cannot be summarized by just a pair of numbers.  To make matters more complicated, lighting is rarely uniform over the entire scene.  So, a global temp/tint white balance is really a ballpark estimate to get you started.

Eric,
Do you think we could expect spectrally characterized cameras and spectral RAW conversion workflow in a near (or not so near), predictable future? In simple words - something that works like an i1Display Pro but has a little bit... more sensels ? Does industry even consider such solution at all? ACR v.7-8-9 or rather 25-56-122?

[madmanchan]

The issue is that spectrum of illumination is generally not known at time of capture.  It is of course possible to provide illumination-specific presets (and profiles), but it's partly a manual process in that the user would have to pick which one to use.  That's also awkward because the user generally won't know (unless he/she has some spectral instrument with them in the field!).  In some special circumstances I think this can be useful, but for general photographic use, I don't think so. 

[Hening Bettermann]

Thanks to all who chimed in.

In the meantime, it has struck me, that my "method", to the degree it could be made precise, would have the same result as using the grey card.

When I started shooting digital, I took a purist approach and set every shot to daylight. This gave trustworthy, if uncontrolled results over a wide range. But some images shot under overcast sky late in the afternoon or early summer evening were absolutely too blue for my memory.

After that, I have relied entirely on the AWB of the camera and then chosen As Shot in the raw converter. The point of departure for my current attempt was that in some cases I discovered that the simple Daylight setting in the raw converter was more conform to my (fresh) memory than the AWB/As Shot combo. I also started wondering how the tint numbers actually were made up, hence my initial question.

Czornyjs answer seemed to make things a little simpler by omitting the tint. Erics post on the contrary points out that matters are even more complex. The whole spectrum of light would need to be considered. "So, a global temp/tint white balance is really a ballpark estimate to get you started." The problem is: started with what?

The concept of mixing in saturation is all new to me, I have no idea how it fits in theoretically. What happens to color saturation, compared to "natural"?
I have no problem making an image look "natural" just by changing the WB - my problem is that it looks "natural" over a wide range, and I would like a little more precision.

- I just tried one image which was shot late afternoon and which I remember I found too blue/green when my memory was fresh. Result:
1-Strange enough, the image does not look unnatural to me today at 5200 K=as shot.
2-It becomes even more "natural" if I increase exposure by 1/2 f-stop - however, the late afternoon mood is lost to some degree.
2-My memory, now 5 years old, accepts WB up to about 7000K, if I insist, that the thing did in fact look green. 7000K is probably pretty close to the temp at shooting time.
3- increasing saturation makes the image more blue/green.
 
No I will not carry additional instruments. Everything that weighs has to be omitted. One problem such instruments would meet is that my shooting "biotope" these days is  woodland, and there is so to speak never an unobstructed view to the light source.

Even if there was, and even if the camera had a built-in spectrometer, there would be one problem left, same as with the grey card: The "AWB" of our brain does not work 100%, and hence, no objective measurement of incident light would have to be followed 100%. But to which %?

For now, I think I will try the following: Shoot with AWB. Then set the raw converter to daylight, say 5200 K. Neglect tint. If the resulting image is in all too strong contrast to my fresh memory, I'll try to adjust temp, staying as close to 5200 K as my memory will permit.

Good light!

[Czornyj]

The issue is that spectrum of illumination is generally not known at time of capture.  It is of course possible to provide illumination-specific presets (and profiles), but it's partly a manual process in that the user would have to pick which one to use.  That's also awkward because the user generally won't know (unless he/she has some spectral instrument with them in the field!).  In some special circumstances I think this can be useful, but for general photographic use, I don't think so.  

Very good point, thanks Eric. I'll get back to a real world

Thanks to all who chimed in.

In the meantime, it has struck me, that my "method", to the degree it could be made precise, would have the same result as using the grey card.

When I started shooting digital, I took a purist approach and set every shot to daylight. This gave trustworthy, if uncontrolled results over a wide range. But some images shot under overcast sky late in the afternoon or early summer evening were absolutely too blue for my memory.

After that, I have relied entirely on the AWB of the camera and then chosen As Shot in the raw converter. The point of departure for my current attempt was that in some cases I discovered that the simple Daylight setting in the raw converter was more conform to my (fresh) memory than the AWB/As Shot combo. I also started wondering how the tint numbers actually were made up, hence my initial question.

Czornyjs answer seemed to make things a little simpler by omitting the tint. Erics post on the contrary points out that matters are even more complex. The whole spectrum of light would need to be considered. "So, a global temp/tint white balance is really a ballpark estimate to get you started." The problem is: started with what?

The concept of mixing in saturation is all new to me, I have no idea how it fits in theoretically. What happens to color saturation, compared to "natural"?
I have no problem making an image look "natural" just by changing the WB - my problem is that it looks "natural" over a wide range, and I would like a little more precision.

- I just tried one image which was shot late afternoon and which I remember I found too blue/green when my memory was fresh. Result:
1-Strange enough, the image does not look unnatural to me today at 5200 K=as shot.
2-It becomes even more "natural" if I increase exposure by 1/2 f-stop - however, the late afternoon mood is lost to some degree.
2-My memory, now 5 years old, accepts WB up to about 7000K, if I insist, that the thing did in fact look green. 7000K is probably pretty close to the temp at shooting time.
3- increasing saturation makes the image more blue/green.
 
No I will not carry additional instruments. Everything that weighs has to be omitted. One problem such instruments would meet is that my shooting "biotope" these days is  woodland, and there is so to speak never an unobstructed view to the light source.

Even if there was, and even if the camera had a built-in spectrometer, there would be one problem left, same as with the grey card: The "AWB" of our brain does not work 100%, and hence, no objective measurement of incident light would have to be followed 100%. But to which %?

For now, I think I will try the following: Shoot with AWB. Then set the raw converter to daylight, say 5200 K. Neglect tint. If the resulting image is in all too strong contrast to my fresh memory, I'll try to adjust temp, staying as close to 5200 K as my memory will permit.

Good light!

Hening, the problem is that even if you know the correct WB value it doesn't really have to result in correct colors. Cameras see colors in a different way than human eyes, and even if we could correct them, they're rendered on a screen or paper that has nothing to do with original viewing conditions. The way we're perceiving colors is amazingly complicated and relative. So I'm afraid you won't find any simple rule here, and you have to rely on your taste, or at least X-Rite CC Passport.

Here's a couple of my favourite essays on this fascinating topic, maybe you'll find something interesting:

http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf

http://www.um.es/phi/aguirao/master/color.pdf

http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf

[digitaldog]

Hening, the problem is that even if you know the correct WB value it doesn't really have to result in correct colors. Cameras see colors in a different way than human eyes, and even if we could correct them, they're rendered on a screen or paper that has nothing to do with original viewing conditions. The way we're perceiving colors is amazingly complicated and relative.

Exactly. Having all the numbers colorimetrically correct doesn’t ensure anyone, especially the image creator, will like the rendering. Its subjective. This goes back to the recent, long and convoluted thread where the term “accurate color” was dismissed by some, (especially me) while we were told its not subjective.

[Hening Bettermann]

Thank you for the links, Marcin. I'll have to digest these first. I have read Karl Lange's paper before and don't quite agree with him.
In the meantime:

> even if you know the correct WB value it doesn't really have to result in correct colors.
 
Well isn't that exactly what I say when I write
>my intention […] is to come a little closer to a "theoretically correct" color. By that I mean the perceived color at shooting time. This is neither the color that would be achieved by using the grey card, nor any arbitrary "pleasing" color.
and
>
Even if there was, and even if the camera had a built-in spectrometer, there would be one problem left, same as with the grey card: The "AWB" of our brain does not work 100%, and hence, no objective measurement of incident light would have to be followed 100%. But to which %?


Andrew, I have followed a part of that long convoluted thread, and I felt that it lacked exactly what I am trying to formulate here: a concept of the PERCEIVED color at shooting time. In that thread the 2 only alternatives seemed to be colorimetric accuracy on the one hand, and completely ad lib subjectivity on the other.

If I am fascinated by some particular color I see in nature, it is THIS color I want to reproduce, not just any arbitrary color. It can be achieved to some degree. The color I get with my home made profiles IS more accurate than RD's default rendering, even though it might be difficult to prove it in court.

I mean all calibrating and profiling and so on has just this purpose: to reduce subjectivity in color rendering,  hasn't it? But I see that the WB has particular challenges.

Good light!

[Czornyj]

I mean all calibrating and profiling and so on has just this purpose: to reduce subjectivity in color rendering,  hasn't it? But I see that the WB has particular challenges.

The purpose of calibrating and profiling is to get reasonably predictable and repeatable results. To some degree it also helps to achieve pleasing and convincing colors from camera, but they'll never really match reality, reproduce the original impression as:
- you see the colors in a different viewing condition, and the change of context changes the relations of colors and spoils everything. It's like trying to reproduce the look of a print on a different paper - you'll never get the look of baryta on a newspaper.
- the camera has different spectral sensitivity than cone cells in the retina, and the spectral sensitivity of cone cells is not constant (even locally). So even if we neutralize the white balance in a part of image there may be other parts that look bad, and so or so some colors may be disorted.
- and so on...

[digitaldog]

I mean all calibrating and profiling and so on has just this purpose: to reduce subjectivity in color rendering,  hasn't it?

In theory maybe. In practice not entirely. The basis of current CMS technology is based on very old concepts, developed long before computers existed and using some very simple assumptions that often don’t jive with how we see (perceive) color. As just one example. Nearly all output profiles assume D50 illuminant. That’s a mighty big assumption. The current technology reduces complex images to solid square colors and don’t take color into context. There are issues with Lab, the color model most of the current technology uses in terms of profiling.

There are color scientists working on color appearance models that could greatly aid in getting closer to your goal. But we’re not there yet.

[Tim Lookingbill]

There are color scientists working on color appearance models that could greatly aid in getting closer to your goal. But we’re not there yet.

Andrew, do you know where these color scientists working on this can be contacted?

I'ld just love to discuss and exchange info on the subject of color appearance as I'm sure would a lot of photographers here at LuLa and elsewhere.

Hening, I took a look at your landscape shots on your website. You really don't like saturation. I see you prefer the 'natural' look. The scenes you captured look very similar to the type of country side around where I live in New Braunfels, Texas.

Do you find shaded brush country that bluish looking? I used to think it did until I started forcing myself to take mental notes how warm white limestone rocks appear under the forest canopy. Tree bark is at least neutral or a bit on the warm side. I get the overly pronounced cyan blue shadows and have to reduce saturation in the blue channel in ACR's HSL panel. Works like a champ.

Please don't take what I'm saying as criticism of your work. Just trying to illustrate the subjectivity of this subject.

[bjanes]

So what I plan to use is not my memory of the colors/the wb themselves. I hope to reconstruct the most accurate WB which is possible after the circumstances using a combination of time of the day and degree of cloudiness. 

The time of day and the degree of cloudiness are a start, but one must realize that daylight is a combination of sunlight, skylight, light scattered from clouds, and reflected light from the ground, vegetation, and surrounding buildings (in the case of a more urban environment). The first three sources are sufficiently close to the Planckian locus that a correlated color temperature describes them reasonablly well (see Handprint) for the graph reproduced from Wyszecki & Stiles, 1982 and reproduced below as well as more information. D55 corresponds to noon sunlight and D55 to noon daylight.



For illustration, sunlight is observable when a shaft of sunlight enters a darkened room through a small window which obstructs most of the skylight. Skylight is shown by light entering a northern facing window. When you mention accurate watch out for the thought police in the form of the DigitalDog . The latter seems to prefer automatic white balance of the camera, but it will fail when the scene contains a predominant strongly saturated color and no neutrals.

As Eric stated, color and tint are a good starting point, but why not merely take a reading from a neutral card illuminated by the same light as the subject (if that is possible).

Regards,

Bill