Empircal Analysis of Auto White Balance


How good is auto white balance? Depending upon whom you ask or where you look, you will get different answers. But it seems that it is nearly impossible to find side-by-side comparisons of how images look with auto white balance versus how they look with white balance set according to the lighting conditions. This tutorial fills that gap.

First let me say that if you do raw image processing, then of course you can fix the white balance after the shot, so in some sense it doesn't really matter what setting you are using. However, I have a remark about that at the end.

On the other hand, not everybody has the ability to do raw image processing, so this topic is particularly important to those with lesser expensive cameras. And while there are still ways to fix up your image with Photoshop or the tool of your choice, it will save you a lot of time and guess work if you get it right in camera.

I'll start out with the facts at the beginning and then jump to an analysis at the end.

How the Experiment is Done

All shots were done with a tripod and a Canon EOS 7d camera (quite a good camera). I pick a subject in a particular lighting condition (shade, cloudy, sunny, flash, white flourescent, etc...). I determine a suitable exposure, shutter speed, and ISO for that subject. I then switch to manual mode and fix those values (exposure, shutter speed, ISO). I take two photos with these fixed values: one with auto white balance and one with white balance matching the shooting conditions. Hopefully I did not shake the tripod much in these shots: if so, I repeat until I get two images that are almost identical other than the white balance processing results.

I then take those two images in jpg form and do no further edits to them other than resize (so that they fit on your screens) and pasting them side-by-side into a single image, so that you can compare the results. I also added textual labels in the upper left corners so you can see which shot is which.

Hence, there are no differences between the two images other than white balance settings.

In all cases below, I put the auto white balance (AWB) on the left, and the white balance shot matching the lighting conditions on the right.

The Data

Lighting condition: cloudy

ISO 100, aperture f7.1, shutter speed 1/30

Lighting condition: shade

ISO 100, aperture f9, shutter speed 1/10

Lighting condition: sunny

ISO 250, aperture f9, shutter speed 1/200

Lighting condition: white fluorescent

ISO 400, aperture f9, shutter speed 1.3

Lighting condition: flash with shoot-thru umbrella (strobist)

ISO 100, aperture f9, shutter speed 1/100

My Analysis

Now the subjective stuff: my analysis.

Cloudy: There is a clear difference in greens and purples. The one shot in auto white balance has "faded" greens, and the purple is also not as strong. In my opinion, the one shot in cloudy white balance wins hands down.

Shade: Similar to cloudy, the biggest differences are in the greens. The one shot in auto white balance has cooler greens than the one shot in shade white balance. The orange is not as strong either. In my opinion, the one shot in shade white balance is the clear winner.

Sunny: Although the differences are more subtle, the result is also similar to cloudy. We get slightly warmer greens and slightly stronger yellows with the white balance set to sunlit conditions.

White Fluorescent: The one shot in auto white balance has brighter reds than the one shot in white fluorescent white balance. In my opinion, this is the one image where auto white balance wins: I prefer the stronger reds.

Flash: The differences between the two are small. Perhaps the reds are slightly stronger in the flash WB shot, but not by much. The other thing to notice is the colour of the plate: the AWB has it white, the flash wb does not. Looking at the real plate right now, it is definitely not white, so I can say the flash wb is more accurate. In my judgment, the flash wb slightly beats AWB here.

My Conclusion

For outdoor lighting, setting your white balance according to the lighting conditions is a big advantage over AWB in terms of the output the camera produces. For indoor lighting, there is no clear advantage.

The analysis so far is based upon only a few images, but I have a lot of experience on this subject from thousands of images. For many years, I shot with a Canon S3 IS which did not have raw image ability. Despite this, I got shots that scored well in competition including at the national level. I depended upon white balance selection for the accuracy of my colours.

I also started out as an underwater photographer, and that was the first thing that gave me the clue to how much a difference it makes. Because if you shoot AWB for underwater photography and if you do not have sophisticated lighting equipment, then all of your images will be extremely washed out. Using under water white balance makes a big, big difference.

Why not just Fix it up in RAW after the Shot?

It is 100% true that if you shoot RAW, you can easily fix the white balance of the image in post processing. Just be aware that results of adjusting white balance in post processing might not be identical to the results you get if you shoot it right in camera. In my experience and according to my own judgment, Canon does a slightly better job at getting the best result when done correctly in camera than Adobe Camera Raw does in post processing. But don't trust my judgment: try it yourself.

But there is Not always Time to Fiddle with Camera Settings!

As a bird photography, I fully appreciate this viewpoint. However, my camera has the ability to register commonly used settings so that I can readily flip to the settings I need with a single turn of a dial. I have one setting for birds in sunlight, one setting for birds on a cloudy day, and one setting for birds in shade. The only difference between the three is ISO and white balance.

Click here to return to Scott Contini's Photography page