Three questions about white balance

[TonyW]

1. If I take a picture then look at the histogram on a the camera screen, is that based just on the raw data or does it take notice of the white balance setting in the camera?

2. When I open a raw file in ACR, I am offered a choice of white balance settings: as shot, auto, daylight, cloudy, etc. None of these, including 'as shot' as I recently found out, present the white balance as the raw data show. How can I see what it looks like without any adjustment of the raw data?

3. How does auto white balance work? Would you expect a big difference between what the camera does and what ACR does when using auto?

[chauncey]

1. That histogram is based on your camera's PP software's jpeg produced image, which in influenced
by any/all of the picture style settings that you have instructed your camera to use in jpeg output.

2. Opening in ACR will not produce any image alterations, assuming that you haven't entered any.

3. Dead on WB requires the use of a http://xritephoto.com/ph_product_overview.aspx?id=1257
used during the shoot...otherwise, it's artistic license. Auto, is jpeg decided.

[Manfred M]

1. The histogram that you see is based on a jpeg. Raw data needs to be assembled into an image before it can be displayed, so whatever presets / defaults you have set for the jpeg on your camera is what the displayed image is based on.

2. No, an image without a white balance setting is meaningless, so a white balance setting needs to be assigned to see it. In the same way, an image file without a colour space is meaningless, so again, one has to choose a colour space before an image can be displayed.

3. Neither the camera manufacturers nor Adobe provide us any information as to how they determine white balance. At a high level, the easiest way to determine the white balance for a scene is to take all the colour data and turn it into a single colour shade (much like what the Photoshop Average filter does) and then determine what the offset each colour channel needs to convert the average colour into a neutral gray and apply that to the whole image.

I strongly suspect that this (at a minimum) is what ACR does when you select auto, however, I can also see it doing this for different ranges of colour values, ranging from white to black and applying a curve calculated from this data. I suspect that the camera's white balance function might perform in a similar manner.

[William W]

1. If I take a picture then look at the histogram on a the camera screen, is that based just on the raw data or does it take notice of the white balance setting in the camera?

From memory you use an EOS 5D MkIII.

On that camera and as far as I am aware on all EOS Series, the histogram is generated by the in camera JPEG file. Even if you shot raw only, the camera will generate a JPEG file that is what is used for the LCD in camera preview. Therefore the “Picture Style” and SOME other settings will have an effect on the Histogram (and the blinkies and also the in camera LCD preview).

I think that High Tone Priority does, but some others don’t – I would have to refer to my detailed notes as I don’t use the histogram either often or in any detailed manner so, for me, if the Histogram is close that’s good enough for me .

I believe that I referenced this exact point in the previous thread about Netball. I understand that this is a new thread and a direct question: trust that’s a detailed and broad enough answer, to get things going.

*

2. When I open a raw file in ACR, I am offered a choice of white balance settings: as shot, auto, daylight, cloudy, etc. None of these, including 'as shot' as I recently found out, present the white balance as the raw data show. How can I see what it looks like without any adjustment of the raw data?

I don’t believe that you can.
'
raw’ data is exactly that ‘raw data’.

In simple terms raw is lots of 1’s and 0’s.

That raw data requires ‘interpretation’ or ‘processing’ to provide an image on screen that we can see.

*

3. How does auto white balance work?

I don’t know the ins and outs but lots of intricate digital functions. Manufacturer’s Proprietary Knowledge, I would think, I don’t think that there is a generic AWB program, but I don know.

*

Would you expect a big difference between what the camera does and what ACR does when using auto?

I would NOT expect a BIG difference, but I have never done any A/B testing.

I expect that there would be two differences, in some scenes:

> firstly the ‘interpretation’ or the ‘bias’ of some parts of the scene being measured by the AWB might vary

> secondly, one AWB program might be intrinsically skewed compared to another – for example in Nikon’s early DSLRS, their AWB seemed to me (and others) intrinsically a bit cooler and Canon’s, by contrast, was a bit warmer.

I can’t comment if this is so for the latest DSLR’s, as I haven’t done any A/B tests on the ore recent cameras. This is like the colour bias that one film might have, when it is contrasted to another film stock - for example, Fuji Professional (printed on Fuji Paper) tended to be a little bit cooler that Kodak Vericolor (printed on Kodak Paper).

HOWEVER - I would expect that ALL AWB programs (and also screen profiling and calibration tools) would interpret, in exactly the same manner, any STANDARD PHOTOGRAPHIC GREY CARD or STANDARD PHOTOGRAPHIC WHITE CARD.

Picking Nits - as a general comment - for all the Digital Cameras (which I have used) - the AWB can be problematic with rich greens and rich purples to keep precise CONSISTANCY across different lighting scenarios; and sometimes even within the same lighting scenario.

For precise work in these conditions (e.g. shooting a Wedding and Bridesmaids in Purple Dresses) I would shoot a Standard Grey Card (and usually Standard Colour Checks also) inside the first frame of each new lighting scenario - and as I mentioned on the other thread, for ease of Post Processing I would often set WB to equal a specific number in degrees Kelvin, so that for each lighting scenario, each image in that batch, would always open in Post Production, at the same reference point.

WW

NB - I am in concert with what wm c boyer and Manfred wrote - we were writing at the same time - (except for this small pedantic bit below about 'dead on')

[William W]

3. Dead on WB requires the use of a http://xritephoto.com/ph_product_overview.aspx?id=1257 used during the shoot...otherwise, it's artistic license. Auto, is jpeg decided.

"Dead on" White Balance requires IN THE FIRST INSTANCE absolute control of the lighting in respect of consistent Colour Temperature and THEN a completely controlled and standardized workflow and also a controlled final viewing area.

"Dead on" white balance is rarely used and, more importantly, is very rarely necessary.

WW

[george013]

3. How does auto white balance work? Would you expect a big difference between what the camera does and what ACR does when using auto?

Both Nikon and Canon have there own RAW converters, both are in-camera as in their out of camera software. Nikon placed that in CaptureNx and ViewNx, Canon in Digital Photo Professional I think. If you've one of them you can compare it with what ACR is doing in any WB-setting. Also the "as shot" might be different.

George

[Mike Buckley]

3. How does auto white balance work?

Nikon cameras have a database of over 10,000 white balance settings stored in them. The camera compares the scene being captured with the database and selects what it "thinks" is the setting that most closely resembles the scene being captured.

Would you expect a big difference between what the camera does and what ACR does when using auto?

On a related issue, I've experienced such big differences between what Nikon's camera and Nikon's raw converter produces when using Auto white balance that I stopped even trying it using the converter. However, that was a long time ago when neither Nikon cameras nor their raw converter had an effective Auto white balance. Their cameras today have a very effective white balance but it has been so long since I tried the converter's setting that I have no idea what it's like. So, I also don't know if a major difference would be produced today.

[Manfred M]

Nikon cameras have a database of over 10,000 white balance settings stored in them. The camera compares the scene being captured with the database and selects what it "thinks" is the setting that most closely resembles the scene being captured.

I don't quite believe that this is correct, Mike. From a technical and implementation standpoint this does not make any real sense. The amount of lookup and comparison for a specific image would be very complex and time consuming. I'm not sure that this would even be something that would be feasible from a technical standpoint.

I rather suspect what really happens is that Nikon has tested their and validated their hardware and algorithm(s) against a large set of test images.

[xpatUSA]

For comparisons of various methods with illustrations, see:

http://www.acad.bg/rismim/itc/sub/ar...er3_1_2012.pdf

A proposed method with comparisons to existing methods, also with illustrations:

http://www.csie.ntu.edu.tw/~fuh/pers...forDigital.pdf

Both of these papers remind me why I rarely use AWB

[TonyW]

Thanks everyone for taking the time to address my nerdy questions. You have given me much to think about and to experiment with. I very much appreciate it.

Given my rather pedantic approach to technical questions, I still have many details that I would like to understand but it would be too tedious to go into them all now.

In most situations, I don't think I have a serious problem with white balance. It is only under difficult circumstances where the lighting is inadequate or too harsh that I seem to get into difficulties. I have always had my camera set on auto white balance and neutral picture style.

[Manfred M]

Tony - the real issue in trying to answer some of your questions is that the answers are quite complex.

What no one has covered is that many lighting situations are actually "mixed lighting", for instance taking a picture of someone standing under a tree in broad daylight or standing beside a red brick wall. No white balance in the world will be able to compensate for those mixed lighting shooting conditions. Something in the image will be "off".

This is why a "best practice" in portrait and product photography has the photographer taking an image of a known target (neutral gray) that is hit by the same light as the subject. That way we can at least get that aspect of the image to look right quite easily.

[Mike Buckley]

I don't quite believe that this is correct, Mike. From a technical and implementation standpoint this does not make any real sense. The amount of lookup and comparison for a specific image would be very complex and time consuming.

I have always thought the same thing since reading about it when I got my first DSLR. At least I remember reading it but my memory could be wrong.

[Mike Buckley]

I have always had my camera set on auto white balance

When I was using my Nikon D80, the auto white balance was so ineffective that I never used it. When I upgraded two generations later to my Nikon D7000 and later purchased a D5100 for my wife, the use of auto white balance is almost always flawless; those two cameras remain permanently set to auto white balance.

[Manfred M]

For comparisons of various methods with illustrations, see:

http://www.acad.bg/rismim/itc/sub/ar...er3_1_2012.pdf

A proposed method with comparisons to existing methods, also with illustrations:

http://www.csie.ntu.edu.tw/~fuh/pers...forDigital.pdf

Both of these papers remind me why I rarely use AWB

Thanks Ted - an interesting read. Tungsten and fluorescent lighting has always be the Achilles heel of AWB in my experience. The other weakness seems to be repeatability. Take a series of shots under the same lighting conditions and each one will show a slightly different colour temperature, when in fact nothing has really changed,

Not an issue in raw, as it is easy to apply a the same correction to the group of images, but when shooting jpeg, it is often better to either use a custom WB or use the same preset for the series of the shots to ensure that all the images in a series look the same from a WB standpoint.

[DanK]

You have received good explanations, and I won't try to add to them. However, I will add a bit of unsolicited practical advice. One reason to shoot raw is so that you don't have to worry about this. As Manfred explained, there is no white balance in the raw file, just raw data. All of the settings converters might use to start, such as the "as shot" setting, are just guesses based on various factors, such as the settings you had in the camera (which have no effect on the raw file). So--here's the nice part--it makes absolutely no difference. Just set the white balance as you want it.

If you are going to be in a situation where WB is questionable, shoot an image with a neutral card, like a whiBal, and use that to find the neutral white balance setting in your raw converter. That will be a true neutral; whatever you get from the camera or software defaults may or may not be. Then adjust to taste if you want.

Re mixed lighting: as Manfred pointed out, this can be a really serious problem because then there isn't one neutral setting. However, I have found that in many cases--e.g., mixed tungsten and flash indoors, or mixed LED and flash indoors--it doesn't matter all that much. You can often get quite acceptable results by shooting a neutral card in a setting like that.

In practice, I pay a little attention (not much) to the WB settings on my camera, just to give me an easier "as shot" starting point. It's also a bit convenience in some cases--e.g., setting a fixed temperature for sunsets. However, it is nothing more than a convenience. It makes no difference at all in terms of what can be produced from the raw file.

[TonyW]

Ted, your references were most interesting and gave some idea of how auto white balance might work (or not as the case might be).

In one of my questions, I was really asking if there is a colour balance in ACR which assumed that the lighting was 'canonical', which is the term used in the first of the your references. I did not follow up their references to find out what 'canonical' lighting is but I would guess that daylight would be the closest.

[xpatUSA]

Ted, your references were most interesting and gave some idea of how auto white balance might work (or not as the case might be).

In one of my questions, I was really asking if there is a colour balance in ACR which assumed that the lighting was 'canonical', which is the term used in the first of the your references. I did not follow up their references to find out what 'canonical' lighting is but I would guess that daylight would be the closest.

Indeed, I would agree that 'daylight' is canonical, being sunlight+skylight at about 5350K. This is pretty close to the CIE 'D55' theoretical illuminant:


The diagram is in the CIELUV space which is more perceptual for color differences than CIE xyY. The black line is the Planckian blackbody color temperature line. As we can see, D55 (and 'daylight' I'll bet) has a slight tint towards green relative to the theoretical 5500K. From this diagram, though, it's pretty clear what ACR's "tint" slider does . . .

Too much information for most, I would think.

[davidedric]

N ice diagram, Ted. I knew that the tint slider works orthogonally to the black body line, but this colour space makes it much clearer, at least to me.

Cheers,

Dave

[TonyW]

Indeed, I would agree that 'daylight' is canonical, being sunlight+skylight at about 5350K. This is pretty close to the CIE 'D55' theoretical illuminant:


The diagram is in the CIELUV space which is more perceptual for color differences than CIE xyY. The black line is the Planckian blackbody color temperature line. As we can see, D55 (and 'daylight' I'll bet) has a slight tint towards green relative to the theoretical 5500K. From this diagram, though, it's pretty clear what ACR's "tint" slider does . . .

Too much information for most, I would think.

No, not too much information. This is a good way to think about it and makes thing clearer. I have had a similar diagram in my mind before.

It is probably true that it is easier to adjust white balance by adjusting the colour temperature and then the tint rather than the individual colours, red green and blue, as photoshop invites you to do, because most of the adjustment needed is along the colour temperature line.

[Ndukes]

Too much information for most, I would think.

That's all really useful. Not too much information.