What is the correct WB?

[dreed]

One of the biggest challenges I have doing sunset photography is white balance. To be able to work consistently, I almost always have my camera set to "daylight" (exception will be indoors) - there's no point in using the camera's histograms to compare exposures if the WB is set to "auto" and thus potentially different for each shot!

One big problem here is "memory". I can't remember exactly what it looked like.

Using LR, the choices are:

* 2000K, Tint -24 (WB from clouds)
* 2850K, Tint 0 (auto)
* 4350K, Tint -47 (WB from yachts)
* 5500K, tint +10 (Daylight)

... all of the other "develop" settings are 0.

[Martin Kristiansen]

You cant remember how it looked? I think that’s the nub of the problem. The idea that there must be an objective way it looked that you are trying to now remember.

In the past we used mostly daylight film for landscape and in the evening it tended to shoot very warm at sunset and pretty cool around mid day. Was common to use a warming haze cutting filter particularly at mid day and that warmed up the blue a little. Thing is with a fixed WB we were then exposed to the variations af ambient WB shift and that created a lot of atmosphere in images that we now tend to correct out when balancing every image to “correct” WB.  You setting daylight for WB is a pretty good way of replicating what we got with film. You don’t have to shoot like that if saving raw files because you can always get back to daylight in LR or ACR or C1 or wherever.

Some photographers, quite a lot on this forum actually, are comfortable with a scientific approach and will try to balance everything to what colour it should be. By that I think they mean what colour it is once the computational imaging system that is your eye and brain is done auto balancing and hdr’ing everything. Then there is the very scientific approach of shooting colour checkers.

Then there is the totally creative approach. How do you want it to look. Different colours and tonal ranges evoke different emotional responses and you se that and other things in the photograph to say something. A scary approach because it relies on intuition, creativity and a belief in what you wish to say.

Photography both scientific and creative. Attracts people that like either approach and even people that like both. Do you want the correct WB or the WB that will most effectively communicates your emotional response to the scene? If you want to do the latter but don’t know what your emotional response was to the scene then you have more important things to do than worry about WB. If the former than you need to spend more time on the Colour management sub forum.

[MauriceRR]

1- Accuracy : expodisc or neutral grey card. It works well when color temperature is between 5000 and 8000Kelvin. (or specific works where neutral WB is mandatory, ex : reproductions artwork)
2- Pleasing mode : no more accurate, but you put your reference white to your taste : often means yeellow/orange cast for under 4500k. Our brain doesn't discount totally the white point, at this stage of temperature. To my opinon, a neutralized sunset is a non sense.

Every AWB algo are just a statistical reply to the problem (they tend to adjust the white point in the purpose to integrate all rgb values from the sensor inside the locus spectrum.)

[32BT]

What's the question?

#2 seems fairly realistic
#4 is how you could sell it these days

You hit upon my main gripe with modern raw processors though. Temperature used to be a good indication of film response, but it has no relevance in our processing logic, especially not when you need to add a second parameter "tint" that has no physical equivalent but merely is a mathematician's solution to a flaw in the modelling.

What they should do is add a colorcast slider with a neutral reference color. This is the way we are used to thinking. This what traditional photography did with "colortemperature" in film. You knew its response and how to compensate for it.

[dreed]

I've got a "grey card" but I'm unconvinced that really helps. Why do I say that?

In a situation like this, I suspect that the right way to use it is to turn away from the scene so that the light coming from the horizon (that will go into the camera) bounces off the card and then into the camera. If I put the grey card between the camera and the horizon, the grey card is not reflecting the "warm" light. Or at least that's how I think. But I've never seen anyone use them like that, so I'm not sure. In a water scene like this, there's scattered light being reflected everywhere off the water - but how much comes from atmospheric scatter? I've never found anyone to properly explain to me how a grey card is expected to work other than "do this". The worst part is the assumption that the light temperature where you are is the same as whatever it is for the object you're photographing. Or am I overthinking this?

[Bart_van_der_Wolf]

You cant remember how it looked? I think that’s the nub of the problem. The idea that there must be an objective way it looked that you are trying to now remember.

That's because of Chromatic adaptation by the Human Visual System (HVS).

FWIW, the approach I use in such cases is to take a reference shot with a WhiBal reference (making it reflect the sunlight, by turning my back to the sun), and use that as a starting-point for White-balancing when Raw converting. Then adjust the Color-balance to something warmer that I feel is appealing. I use Capture One Pro as Raw converter, and it has excellent tools for that. One can even apply a cooler Color-balance to shadows, and a warmer Color-balance to highlights and/or mid-tones.

Cheers,
Bart

[Bart_van_der_Wolf]

I've got a "grey card" but I'm unconvinced that really helps. Why do I say that?

In a situation like this, I suspect that the right way to use it is to turn away from the scene so that the light coming from the horizon (that will go into the camera) bounces off the card and then into the camera. If I put the grey card between the camera and the horizon, the grey card is not reflecting the "warm" light. Or at least that's how I think.

And you are correct.

There are other (sometimes subtle) effects, like when under a tree canopy that filters/reflects towards green light, there is also potentially a lot of ambient light coming from reflected light from the soil, not only incident sun/cloud-light. So that's where Creativity comes in.

Cheers,
Bart

[Martin Kristiansen]

That's because of Chromatic adaptation by the Human Visual System (HVS).

FWIW, the approach I use in such cases is to take a reference shot with a WhiBal reference (making it reflect the sunlight, by turning my back to the sun), and use that as a starting-point for White-balancing when Raw converting. Then adjust the Color-balance to something warmer that I feel is appealing. I use Capture One Pro as Raw converter, and it has excellent tools for that. One can even apply a cooler Color-balance to shadows, and a warmer Color-balance to highlights and/or mid-tones.

Cheers,
Bart

I agree with all that but use a slightly different approach which I don’t think is better or worse and is unimportant to this discussion.

I like the different colour balance idea in different areas Bart and it is this approach that is one of the reasons I prefer C1 to LR for my personal work. I find LR split toning clumsy by comparison.

Out of interest I am also inclined to desaturate in shadows as human vision loses colour discrimination under low light conditions and I think brightly coloured shadows can look garish and unnatural. I try to be subtle with it of course but a uniform saturation across a broad tonal range lacks the subtlety of human vision and takes some of the mystery out of shadow areas.

[kimballistic]

The correct WB is whatever feels authentic to your memory of what you saw and felt.  The more your chosen WB can evoke what you felt while being there, the better.

Shooting grey cards and other WB tools will eliminate the light's color cast for when you want accurate subject color independent of lighting conditions.

Now consider beautiful light like your example sunset, where you're trying to capture the color of the light, not eliminate it.  There's no procedure or tool to automate this.  You just need to feel it out and use your artistic judgement.

[digitaldog]

The correct WB is whatever feels authentic to your memory of what you saw and felt.  The more your chosen WB can evoke what you felt while being there, the better.

Shooting grey cards and other WB tools will eliminate the light's color cast for when you want accurate subject color independent of lighting conditions.

Now consider beautiful light like your example sunset, where you're trying to capture the color of the light, not eliminate it.  There's no procedure or tool to automate this.  You just need to feel it out and use your artistic judgement.

Exactly!
This has little if anything to really do with color accuracy and everything to do with producing pleasing color. The correct WB is the one the image creator produces while rendering the image as they desire. A HUGE part of the process, science and art of photography as outlined so well in this article by Karl Lang:
http://www.lumita.com/site_media/work/whitepapers/files/pscs3_rendering_image.pdf

You want to talk accurate color? You need to start handling scene referred color which is often rather ugly:
http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf

You want to talk accurate color, you need a way to measure and define color accuracy:

Delta-E and color accuracy

In this 7 minute video I'll cover: What is Delta-E and how we use it to evaluate color differences. Color Accuracy: what it really means, how we measure it using ColorThink Pro and BableColor CT&A. This is an edited subset of a video covering RGB working spaces from raw data (sRGB urban legend Part 1).

Low Rez: https://www.youtube.com/watch?v=Jy0BD5aRV9s&feature=youtu.be
High Rez: http://digitaldog.net/files/Delta-E%20and%20Color%20Accuracy%20Video.mp4

You want to talk WB and RGB value neutrality, you can discuss WB cards and software that sets WB. That rendering using such cards may or may not produce the rendering the creator desires. Useful to get to a starting point, maybe (rarely for my renderings) is it a one click and I'm done affair; I season to taste.

You want to talk remembering the colors you 'saw' at the scene and them matching them on a display or print? Good luck with that. It's so utterly subjective. IF you feel you matched the scene on a print or display (without regard to dynamic range, Chromatic adaptation etc), you're done.

Speaking of Chromatic adaptation, our friends at Chromix just published a very good article on the subject:
The Ideal Room with a View

[32BT]

I wonder in how far this is true. If we introduce a slider called "temperature" in a photographer's colortoolbox (i.e. raw processor), is it then supposed to mimic traditional behavior as understood by a traditional photographer? Or is it just some random parameter in the multitude of unexplained parameters also available to adjust the image by fingerspitzen gefuhl?

In the former case, should the slider represent scene lighting? Or sensor response?

[digitaldog]

I wonder in how far this is true. If we introduce a slider called "temperature" in a photographer's colortoolbox (i.e. raw processor), is it then supposed to mimic traditional behavior as understood by a traditional photographer? Or is it just some random parameter in the multitude of unexplained parameters also available to adjust the image by fingerspitzen gefuhl?

In the former case, should the slider represent scene lighting? Or sensor response?

Well if the temperature is based on Kelvin, as we see in LR/ACR (at least textually), useful to know that ANY Kelvin value defines a large range of possible colors:





See that line running from e to f (lines of Correlated Color temperature)? ANY color can be considered 5000K.
The question becomes, which color do you visually prefer, not what number its assigned.

[32BT]

Well if the temperature is based on Kelvin, as we see in LR/ACR (at least textually), useful to know that ANY Kelvin value defines a large range of possible colors:





See that line running from e to f (lines of Correlated Color temperature)? ANY color can be considered 5000K.
The question becomes, which color do you visually prefer, not what number its assigned.

Yes, exactly, which is why they introduced a second parameter "tint", but that unfortunately creates the mix up between references. The temperature slider is meant to reference the capture temperature response. The tint is then used to adjust undesired reproduction. This, according to my not so humble opinion, especially on this subject, is the source of all confusion. The OP is the prime example, and the answers show that people misunderstand the concept. Hence, it turns into a "move it any way you please to get the desired result", which comes dangerously close to "random" as far as I'm concerned.

[digitaldog]

Tint represents magenta to green "axis" and varies as well.
The identical Kelvin or Temp or Tint value in one raw processor will very likely create different color appearance than the other. From the same raw data. Why then edit using a number rather than by what you see? And of course the calibration of the display comes into play.
It's subjective. It really has nothing to do with accurate color.
 

[32BT]

Tint represents magenta to green "axis" and varies as well.
The identical Kelvin or Temp or Tint value in one raw processor will very likely create different color appearance than the other. From the same raw data. Why then edit using a number rather than by what you see? And of course the calibration of the display comes into play.
It's subjective. It really has nothing to do with accurate color.

Not in its current implementation, no. That's why I am a proponent of a different WB paradigm in raw converters: colorcast and colorcastreduction. That way it becomes far more intuitively clear and procedurally objectified what is being adjusted by the slider(s).

In addition, it can become more accurate by introducing a single colorresponse whitepoint for the state of the sensor in the captured rawdata. Suppose a sensor is calibrated to capture D65 as neutral, then it could simply communicate that wp in the rawdata. The rawconverter can then use that as a reference to introduce the scene colorcast. A slider called colorcast can then simply adjust between completely corrected neutral gray, to completely uncorrected colorcast. If you allow overshoot, you have the creative adjustment we are seeking here: partially remove colorcast, or overshoot to introduce even more colorcast, or go funky and invert the colorcast (negative values). The colorcast itself can be added as the "second parameter" required for full WB control. Adjust the colorcast by clicking a desired gray, for example.

Note that it then becomes a more sensible adjustment: 0% could mean "as captured", 100% could mean "completely neutral", (or vise versa) anything in between is exactly that. 

Moreover: it is then also clear that the actual colorresponse of the sensor in relation to the spectral characteristics of the scenelighting, needs to be separated from the WB sliders.

[digitaldog]

Not in its current implementation, no what? The OP is using LR.

[DP]

A slider called colorcast can then simply adjust between completely corrected neutral gray, to completely uncorrected colorcast.

well, unfortunately there are at least 3 different dyes used in CFA toppings... so one slider can't do it 

[DP]

Suppose a sensor is calibrated to capture D65 as neutral, then it could simply communicate that wp in the rawdata.

a lot of cameras communicate in raw data WB settings as raw channel "multipliers" (or rather numbers that software can use to calculate multipliers) for a lot of illuminants, for example

A7R2
0x0264 WB RGB Levels                 : 602 256 394
0x7820 WB RGB Levels Daylight        : 2444 1024 1600
0x7821 WB RGB Levels Cloudy          : 2644 1024 1468
0x7822 WB RGB Levels Tungsten        : 1496 1024 2868
0x7823 WB RGB Levels Flash           : 2696 1024 1444
0x7824 WB RGB Levels 4500K           : 2200 1024 1800
0x7825 WB RGB Levels Shade           : 2916 1024 1320
0x7826 WB RGB Levels Fluorescent     : 2228 1024 2256
0x7827 WB RGB Levels Fluorescent P1  : 2368 1024 1680
0x7828 WB RGB Levels Fluorescent P2  : 2660 1024 1552
0x7829 WB RGB Levels Fluorescent M1  : 1764 1024 2760
0x782a WB RGB Levels 8500K           : 3072 1024 1252
0x782b WB RGB Levels 6000K           : 2620 1024 1480
0x782c WB RGB Levels 3200K           : 1676 1024 2492
0x782d WB RGB Levels 2500K           : 1300 1024 3456

Fuji X-H1
0xf00c WB GRB Levels Standard  : 302 379 845 17 302 644 492 21 ----- #17 = StdA, #21 = D65
0xf00d WB GRB Levels Auto      : 302 710 536

and no - no consumer level camera is calibrated (CFA spectral transmission) for D65 or any other illuminant resembling anything naturally occuring as it is UniWB illuminant.

[32BT]

well, unfortunately there are at least 3 different dyes used in CFA toppings... so one slider can't do it 

I said one slider plus a colorcast (= color, stripped of luminance)

That gives you 3 degrees of freedom, except the parameters at least have some procedural significance in the thinking of a photographer or image editor.

[32BT]

a lot of cameras communicate in raw data WB settings as raw channel "multipliers" (or rather numbers that software can use to calculate multipliers) for a lot of illuminants, for example

Yes, and they are completely irrelevant as far as whitepoint is concerned. The only thing we need to know is the whitepoint of the sensor at the time of capture. Nothing to do with the scene lighting. Merely its electronic state.

and no - no consumer level camera is calibrated (CFA spectral transmission) for D65 or any other illuminant resembling anything naturally occuring as it is UniWB illuminant.

Of course not. They obviously have a CCT only, and D65 was merely taken as example. What can happen (but not likely) is that the sensor is calibrated towards D65. The dyes are chosen and electronics set so the sensor behaves as if it has a CCT of ~6504K or thereabouts.

If we take the multipliers from your A7 as an example, and we deem equal R and B channel as the likely wp, then the CCT of the A7 is somewhere between 3200K and 6000K according to that list.

It is that particular setting that we are interested in. No other information is relevant, as we are going to adjust the parameters (and thus the multipliers) to our personal taste anyway.

Note also that the list provided in the raw data, is very odd from a procedural perspective: the camera obviously doesn't know which lighting it is capturing, and therefore it knows nothing about the temperature, or the required colorresponse. It could be capturing a sunset with low wp temperature, but requiring a daylight colorresponse. Or it could be capturing an incandescent scene with low temperature, requiring an incandescent colorresponse. The camera doesn't know, shouldn't care, and therefore, shouldn't communicate the info.

You could potentially add maybe a spectrophotometer in a camera in some future, so that it can communicate the info as a spectrum, but that is completely beside the point. (One row of spectral dyes though would be really cool, no? Add a white transparant cup to your lens, capture the spectrum...)