How do you set white balance?

[lunaticitizen]

I usually use AWB (Auto White Balance) and sometimes just adjust it on post if the result didn't feel 'right'.

Recently I read articles about how to set custom white balance using a reference card and began experimenting. Here are some of the articles I read:

• Schewe, J 2004, A Digital Workflow for Raw Processing Part Three: White Balance,<https://www.adobe.com/digitalimag/pdfs/ps_workflow_sec3.pdf>
• 18% Gray Card verses a White Balance Card, <https://www.xrite.com/service-support/18graycardversesawhitebalancecard>

One interesting thing I learned from the above articles is that apparently you're not supposed to use a gray card to do white balancing.

Here is what the Adobe whitepaper said:

To be clear, this tool is designed to be used on light, nonspecular neutral tones, such as the second brightest patch on the ColorChecker. It is not designed to be used on a gray card sample. A gray card's tonality is too far down the luminance scale. Because of the lower tone value, there may be noise that can affect the resulting accuracy of the white balance.

And here is what the X-Rite article said:

While you can use the 18% Gray to do a white balance it is not something that we suggest as the results are balancing to the gray not to the neutral white.

Anyhoo, armed with a ColorChecker and a WhiBal card, I took a picture of a model figure and set the white balance on post. Here is the result.



They are all so similar. I had expected that at least the AWB measurement would have been inaccurate but the results were so close. In the end, the experiment was not very interesting

[davidedric]

" In the end, the experiment was not very interesting "

No, I don't agree. I think it's helpful to report the results of well set up experiments whatever the outcomes (in fact, it can be a problem when "uninteresting" or negative results don't get reported)

Thanks for doing the work.

[DanK]

In my experience, the adequacy of AWB in my cameras depends on the lighting and the mix of colors in the scene. It works very well much of the time but can fail under some conditions.

I'm not one to argue with Jeff Schewe, but what he said doesn't make much sense to me. Here's the quote:

It is not designed to be used on a gray card sample. A gray card's tonality is too far down the luminance scale. Because of the lower tone value, there may be noise that can affect the resulting accuracy of the white balance.

How bright a neutral card appears in the image you are processing depends on how you expose the image, not on the characteristics of the card. If you expose a card so that it is near the right side of the histogram, there is very little noise. It seems to me that noise is only an issue if you expose so low that color noise becomes an issue. Perhaps I misunderstand him.

I think the x-Rite comment is ambiguous and is referring to something else entirely:

While you can use the 18% Gray to do a white balance it is not something that we suggest as the results are balancing to the gray not to the neutral white.

But adjusting WB in post doesn't adjust for luminosity. It adjusts for hue. I think this quote be referring to gray cards that are designed for exposure and are not spectrally neutral and are therefore not appropriate for white balance. Cards like a whiBal are designed for the latter purpose, as you can see from the graph here.

A color checker clearly allows more finely grained adjustment. But if you are content to adjust hue on the standard two axes in the white balance adjustment--blue to yellow, green to magenta--I think a well exposed whiBal should do the trick.

Again, I can't claim the expertise of either Jeff Schewe or a spokesperson for Pantone, but my experience is that setting white balance using a well-exposed whiBal works very well. Your experimental results seem consistent with this. I haven't bothered setting a custom white balance for many years.

[Manfred M]

Carefully said, I understand where Schewe's and the Pantone article are coming from, but in general, I find it makes little difference as the readings from a gray card, a ColorChecker Passport or a WhiteBal card are only a decent starting point. I virtually always tweak my white balance by eye because these techniques only deal with the technical aspect on how an image appears and ignore two other key factors, the physiological (i.e. how our visual system - eyes and brain) interpret colours and the psychological aspects of colour (how we perceive an image that is warm toned versus cool toned).

I would also agree that there are weaknesses in my approach based on personal (learned) biases and issues with my visual system (a male in his mid-60's will not perceive colours the same way as a younger person).

Take a look at the colour readings, using the Photoshop color sampler tool on the "neutral" swatches of my of my ColorChecker Passport. Ideally, the RGB colour values should all be the same for each colour swatch, but they are not. Perhaps I should replace the tool, but I noticed this effect when it was brand new. More likely the light hitting each of the squares was a touch different due to the shooting environment.

In theory, if you are not adjusting the "exposure" of an image any target will get you close. Sometimes I want the image to be slightly warmer (as I often do in my portraiture work) and I will bias slightly towards the reds and yellows. Sometimes I want a cool tone and will bias towards the blues and cyans.

[xpatUSA]

"How do you set white balance?"

It depends on the purpose, the scene, and the camera: I have no fixed method.

When I shoot full spectrum+IR or just IR, WB is usually irrelevant because I work off the raw data which is unaffected by the WB setting.

On the rare occasions that I use Custom WB, I shoot the white side of a Kodak R27 card and I shoot it out-of-focus to get an homogeneous capture.

Indoors, I have a Colorchecker card on the wall and I'll color-balance on the mid-gray patch #22. If going to extremes, I'll adjust the white and black patches in Levels and re-check mid-gray. Iteration is sometimes needed.

Somewhere, I have an article which compares the image coloration from color balancing on a scene shot at the wrong WB versus color balancing on a scene shot at the correct WB - it shows significant differences. If anyone is interested, I'll have a look for it.

[lunaticitizen]

Dan,

Yes, I had a hunch that the accuracy of AWB depends on the scenes and their environment. Maybe for this experiment AWB works well with my test lab aka my living room.

My original motivation was to test the dedicated WB sensor of my camera to see if it's good enough, and to compare it with my old camera with no dedicated WB sensor. Unfortunately the old camera is with my wife and she's on a business trip somewhere; so the next test will have to wait.

Re gray card, my understanding is that both Adobe and X-Rite talked about the same thing which is the 18% gray card and why it shouldn't be used as a replacement of white balance card. Honestly I thought WB card was the same as 18% gray card so it was a bit confusing to me.

And yes, as you can see from my little experiment the results from the ColorChecker WB card and WhiBal card were similar. Actually I'm not quite sure on how to interpret the numbers. Can we say the difference of temperature (4500K vs 4400K, ≒ 2%) is insignificant? How about the tint (31 vs 28, ≒ 10%)? All I understand is that they all look 'similar'.

[lunaticitizen]

Manfred,

I'm not too bothered with accuracy, either. I'm mainly a JPEG shooter. That's why I've always used AWB. On the rare occasion I do post-processing, I don't even bother to adjust the WB. The AWB is good enough, although sometimes I use "Auto" too.

Take a look at the colour readings, using the Photoshop color sampler tool on the "neutral" swatches of my of my ColorChecker Passport. Ideally, the RGB colour values should all be the same for each colour swatch, but they are not. Perhaps I should replace the tool, but I noticed this effect when it was brand new. More likely the light hitting each of the squares was a touch different due to the shooting environment.

Hmm this is interesting, I'll try that with my Passport too, later.

[lunaticitizen]

Ted,

Your method is kinda complicated, I'm not sure I'll ever have the will and motivation to try it

Please do share the article.

[lunaticitizen]

" In the end, the experiment was not very interesting "

No, I don't agree. I think it's helpful to report the results of well set up experiments whatever the outcomes (in fact, it can be a problem when "uninteresting" or negative results don't get reported)

Thanks for doing the work.

Dave,

I guess I just needed a validation that I didn't waste my money for a ColorChecker WB card and a WhiBal card

[Manfred M]

s from the ColorChecker WB card and WhiBal card were similar. Actually I'm not quite sure on how to interpret the numbers. Can we say the difference of temperature (4500K vs 4400K, ≒ 2%) is insignificant? How about the tint (31 vs 28, ≒ 10%)?

In general, the higher the colour temperature, the less likely we are able to notice any difference between two different values. You will notice the difference between 3000K and 3300K more easily whereas a 5000K and 5500K will look close. Both are 10% different.

My understanding is that this is due to human physiology, i.e the way our visual system interprets colour temperature. Get into the bright north light at 10000K or higher, the impact is even less noticeable.

[DanK]

Re gray card, my understanding is that both Adobe and X-Rite talked about the same thing which is the 18% gray card and why it shouldn't be used as a replacement of white balance card. Honestly I thought WB card was the same as 18% gray card so it was a bit confusing to me.

There are two separate issues here. First, in the past, people often used 18% gray cards to set exposure. These cards did not need to be spectrally neutral. To set white balance, the cards have to be spectrally neutral. The whiBal is.

Second, I think Schewe's argument is different and has to do with luminosity. He wrote:

It is not designed to be used on a gray card sample. A gray card's tonality is too far down the luminance scale. Because of the lower tone value, there may be noise that can affect the resulting accuracy of the white balance.

Color noise is an issue if a white balance source is badly underexposed, even if it is spectrally neutral, as you might put the eyedropper on a pixel that isn't neutral because of color noise. However, if the card is well exposed, this isn't an issue.

With respect to WB, the main difference between shooting JPEG and raw is that JPEG makes it harder to make corrections if WB is far off. Many years ago, when I first got a digital camera and was shooting JPEGs, I had some bad WB problems that I couldn't fully correct. A cousin of mine who teaches photography at the university level just smiled when I explained the problem and said, "that's one reason I shoot raw."

[xpatUSA]

Ted,

Your method is kinda complicated, I'm not sure I'll ever have the will and motivation to try it

I have no fixed method

Which one?

Please do share the article.

I'll have a look for it ...

[lunaticitizen]

In general, the higher the colour temperature, the less likely we are able to notice any difference between two different values. You will notice the difference between 3000K and 3300K more easily whereas a 5000K and 5500K will look close. Both are 10% different.

My understanding is that this is due to human physiology, i.e the way our visual system interprets colour temperature. Get into the bright north light at 10000K or higher, the impact is even less noticeable.

Yes, I just tried that and you were right. Thanks.

[lunaticitizen]

There are two separate issues here. First, in the past, people often used 18% gray cards to set exposure. These cards did not need to be spectrally neutral. To set white balance, the cards have to be spectrally neutral. The whiBal is.

Yes, I know that 18% gray cards are used to set exposure. But how did you know that these cards didn't need to be spectrally neutral? Why is that?

I'm not really familiar with the term "spectrally neutral". Does it mean that the cards "reflects equal amounts of red, blue and green in all types of light conditions"? I found the definition on this product page of ColorChecker 18% Gray Balance card. Does it mean that it is an unnecessary feature for an 18% gray card?

For what it's worth, according to this ColorChecker product catalog you can use their 18% gray balance card to set white balance using software, which I think is inconsistent with the support page I linked before.

Color noise is an issue if a white balance source is badly underexposed, even if it is spectrally neutral, as you might put the eyedropper on a pixel that isn't neutral because of color noise. However, if the card is well exposed, this isn't an issue.

OK, that makes sense. I might do some experiments about this later.

With respect to WB, the main difference between shooting JPEG and raw is that JPEG makes it harder to make corrections if WB is far off. Many years ago, when I first got a digital camera and was shooting JPEGs, I had some bad WB problems that I couldn't fully correct. A cousin of mine who teaches photography at the university level just smiled when I explained the problem and said, "that's one reason I shoot raw."

Understood. I thought I'd just set the custom white balance on camera, if I can ever be bothered to use the card outdoor

[lunaticitizen]

Which one?

Imprecise language.. I meant having to choose the method depending "on the purpose, the scene, and the camera" might be too much for me.

[DanK]

Yes, I know that 18% gray cards are used to set exposure. But how did you know that these cards didn't need to be spectrally neutral? Why is that?

Because they weren't used to set white balance. They were a common tool back in the film days, and film cameras don't have WB settings. That's why gray cards then didn't need to be precisely spectrally neutral.

I'm not really familiar with the term "spectrally neutral". Does it mean that the cards "reflects equal amounts of red, blue and green

Essentially, yes. See the explanation and graph on the whiBal cover page, http://michaeltapesdesign.com/whibal.html.

I thought I'd just set the custom white balance on camera, if I can ever be bothered to use the card outdoor

To be accurate, custom white balance must be set for every lighting condition. A lot more bother than taking a shot with a whiBal in it and adjusting later in post, IMHO.

Over the years, I've developed a simple system. In situtations where AWB is likely to be close, I don't bother with a card and adjust to taste if necessary in post. If I have any concern that it will be hard to do that, I take one shot in that particular lighting condition with a whiBal, adjust WB on that shot in Lightroom, and then sync the WB setting to all of the other images in that set, which you can do with one operation.

The one circumstance where that often fails is doing candid shots of kids, which I do a lot. These are often in mixed lighting, and kids move too fast to allow a second shot with a whiBal. This is also a situation where AWB will give quite different estimates from one shot to the next. So I end up doing a fair amount of fussing in post. However, custom WB would be impossible in these circumstances as well, as the CWB for one shot would be wrong for the next.

[Manfred M]

Carefully said, the problem is not so much gray card versus other cards, but rather colour neutrality. Unless the target says it is spectrally neutral it may not be, I have a number of gray targets that are both spectrally neutral and 18% gray (assuming that the card suppliers are being honest). If the supplier is reputable; companies like WhiteBal, DataColor, X-rite, etc. I believe them because that is the business that they are in (and they use stable inks and substrates). An inexpensive card from an unknown source on Amazon, eBay etc, you are taking your chances.

That being said, these solutions may be good enough for the work that you do. Unless you are doing commercial work and are trying to get the exact Pantone colours on a client's logo, close enough is generally good enough. I have known a number of wedding / portrait and event photographers who do shoot JPEG out of necessity. None of the good ones use AWB because the camera will not give them a consistent WB and the images will have a different WB across the shots that they have taken. The ones I have spoken with do one of three things; shoot a present (daylight, for instance), shoot at a constant manually set colour temperature (set the camera to 5600K and use that setting) or do a custom white balance and stick with that unless the lighting really changes. One trick I did learn from a commercial event photographer is that if you use your camera's custom WB function and use the scene you are shooting, the white balance you get is often good enough.

The other question, in my view, is to determine what is good enough. If you look at your camera spec, you will probably find that the range of AWB is more limited than setting a manual WB or using one of the presets, especially at the warmer and cooler ends.

Finally, if there is a "white", "black" or "gray" area in your image, you can often pull off a decent WB from them in post. A car tire is usually black, street signs are often white (as are paint lines on a road) or wedding dresses, concrete or asphalt are often fairly neutral. It may not satisfy everyone, but are often good enough for a starting point.

[xpatUSA]

Ted,

Please do share the article.

Here y'are, Leo

https://x3magazine.com/not-all-white...created-equal/

It's quite interesting with a good few illustrations ...

[Manfred M]

Here y'are, Leo

https://x3magazine.com/not-all-white...created-equal/

It's quite interesting with a good few illustrations ...

It seems that the Sigma software is set up slightly differently than how the Bayer sensor based software does this. I don't know if this is necessary or that someone went out of their way to make things more complicated than necessary.

I know that Sigma uses Bayer sensors in some of their cameras. Is this approach used for those sensors as well, or is this something necessary for Foveon sensors? At the end of the day, both types of cameras deal with red, green and blue data. The de-mosaicing step is not necessary for the Foveon based cameras, so why the additional complication of having to choose the type of WB?

The process outlined in the article bears a strong resemblance to what I went through in the colour "wet" darkroom when making subtractive process colour prints. The film type (daylight or tungsten) was determined the starting point and making minor tweaks in the printing process with single or pairs of yellow, magenta and cyan filters is how colour corrections were made.

[lunaticitizen]

Because they weren't used to set white balance. They were a common tool back in the film days, and film cameras don't have WB settings. That's why gray cards then didn't need to be precisely spectrally neutral.

OK. I'm not a film guy, so thanks for the history lesson.

Essentially, yes. See the explanation and graph on the whiBal cover page, http://michaeltapesdesign.com/whibal.html.

Alright. I also saw the page, but I wasn't so sure if they're authoritative enough to be cited here. I own their WhiBal card, though.

To be accurate, custom white balance must be set for every lighting condition. A lot more bother than taking a shot with a whiBal in it and adjusting later in post, IMHO.

Over the years, I've developed a simple system. In situtations where AWB is likely to be close, I don't bother with a card and adjust to taste if necessary in post. If I have any concern that it will be hard to do that, I take one shot in that particular lighting condition with a whiBal, adjust WB on that shot in Lightroom, and then sync the WB setting to all of the other images in that set, which you can do with one operation.

Yeah, I understand that. I don't do much post-processing work, though. That's why I'd rather get the WB right in the first place. If the WB sensor of my camera is accurate enough it's even better. This was one of the reasons I bought this camera. The ability to shoot flash with electronic shutter is another reason.

The one circumstance where that often fails is doing candid shots of kids, which I do a lot. These are often in mixed lighting, and kids move too fast to allow a second shot with a whiBal. This is also a situation where AWB will give quite different estimates from one shot to the next. So I end up doing a fair amount of fussing in post. However, custom WB would be impossible in these circumstances as well, as the CWB for one shot would be wrong for the next.

I also got the WB wrong a lot when I shoot indoor. Mostly I realized I got it wrong after I uploaded the SOOC JPEG file to Flickr and deleted the raw file