What does "Auto White Balance" Mean ?

daveleo

what?
Local time
12:01 AM
Joined
Oct 20, 2010
Messages
3,692
Location
People's Republic of Mass.
I used an 81A glass filter on my Fuji XE1/35mm setup and made some pictures, using the auto white balance setting. My belief was that the camera's algorithm would "compensate" for the 81A tint (I only used it to protect the lens glass), but it didn't work that way. A picture with the filter on is slightly warmer than with no filter.
So I am wondering . . . why does "auto white balance" mean ?

Thanks for your replies.
 
The software in the camera analyzes the photo and decides what is supposed to be white and based on that analysis it adjusts the areas that looks like it should be white based on the light temp. All other colors are adjusted in line with the white reading. It has limitations however and will not be able to compensate for extreme conditions. Some cameras are better than others. My old D90 had terrible auto WB but my D610 has a very smart computer and as such adjusts much better...lol
 
^ I kind of knew that but what I don't get is (thinking about this very slightly warm 81A filter) ..... if *all* the colors are equally displaced very slightly toward the filter color, the "what is white ? " algorithm should easily compensate for this.\

Obviously . . . it does not work, but what I am hoping to learn is WHY it doesn't work, even in this minor color shift situation.

FWIW, I typically shoot RAF in the Fuji and set degK for white balance, and convert to JPG in the camera (using "camera settings"). I understand how THAT (setting degK) would show the effect of the filter, but don't get why auto-WB does not compensate for the color shift. I was thinking that auto-WB would float the degK to get what it is looking for.

EDIT: I think I got it. The camera does some kind of weighted average of pixel colors. Then adjusts the color temperature so that the weighted average matches some preset value (that has a broad tolerance), and uses that color temperature to create the camera JPG. Apparently the preset tolerance is broad enough that it doesn't get exactly the same answer if you do something like use a mild color filter (like an 81A).
 
I'm not sure why you'd put a warming filter on in the hope that the camera would compensate for it?

If it did compensate for it, you'd have the same result as if you had no filter surely?
 
I'm not sure why you'd put a warming filter on in the hope that the camera would compensate for it?

If it did compensate for it, you'd have the same result as if you had no filter surely?

I put it on to protect the lens, not to adjust the color. It was only temporary and I was surprised at what I saw, which raised this question.
 
Automatic White Balancing in Digital Photography

Quote from the below linked PDF document: "The automatic white balancing (AWB) algorithm employed in the camera imaging pipeline is thus critical to the color appearance of digital pictures. This chapter is devoted to a study of such algorithms commonly used in digital photography."

You can find the full document here:

https://courses.cs.washington.edu/courses/cse467/08au/labs/l5/whiteBalance.pdf
 
Modern cameras are a bit smarter than just trying to neutralize all colour casts. Maybe too smart for their own good. I think most cameras nowadays will try to analyze the scene, so if the camera thinks you've taken a picture of foliage it won't try to get rid of all the green. Or in your case it might think you're taking pictures of an evening/sunset kind of scenario and therefore keep the warm tone.
 
Thanks for the new comments.
Thanks shimokita for that link, it was very interesting. Obviously there is a variety of white balance methods and algorithms - the color output variations (design-to-design) are not simply due to how well the camera executes a standardized method.
The next hot in-camera software will offer menu options for which WB method you want to use for the RAF-JPG conversion. 😀

In fact, I just checked the XE1, and you *can* select WB options for in-camera conversion (degK and color shift). And of course you can do any number of those for each image. Too cool.
 
It's a shame it varies so much from one brand to another. My Fujis are pretty bad at AWB in very warm light, the Sony a6000 on the other hand has great AWB in low color temp. light. I shoot only RAW so it does not matter much, but the difference is pretty dramatic.
 
I understand your surprise, and I might even be happy that my camera doesn't make colored filters obsolete automatically (not that there's any point on a digital anyway, to my knowledge)! Fun thread to read.
 
As far as this particular scenario goes the easy "fix" would be to go into the X-E1's white balance settings where it lets you set WB Shift for AWB, and bias it towards the blue end of the spectrum to offset for the filter. It should be consistently a little too warm, so moving it a couple notches cooler should give you roughly an approximation of what you want.

That's assuming you wanted a "fix" as opposed to just being intellectually curious how the Fuji's AWB algorithm actually works under the hood 😀
 
Hey Dave,

just to add another factor to the mix: some newer cameras have external ambient light sensors, the reading of which is included in the WB computation. That sensor has no idea about what filter you use. I think the X100 has an ambient light sensor, the M240 even two (one for LCD panel, another that affects at least the EXIF value, if not the white balance). Not sure about the XE1.

Roland.
 
Roland, you had me scrambling to the cabinet for my XE1 there, while reminding me of the Konica Hexar's external meter ..... but no.... the XE1 uses TTL metering.

@jloden . . . this is not really a "problem" that I am trying to deal with. Just trying to understand what my camera thinks it's doing 😀
 
@jloden . . . this is not really a "problem" that I am trying to deal with. Just trying to understand what my camera thinks it's doing 😀

Gotcha... I wasn't sure from the original post whether you were looking for a workaround or just wondering 🙂

I'm sure this behavior is easily explicable with more internal knowledge of which AWB algorithm Fuji is using. However, I'm equally sure that's proprietary info not available to the public for obvious reasons. Not having read shimokita's linked article on common WB algorithms, just as a wild guess two possibilities that come to mind:

1) The algorithm is optimized for Fuji's lenses and expected light transmission characteristics, and gets thrown off by the filter. Some legacy or adapted lenses add a cool or warm cast to a scene so it doesn't seem too unlikely a possibility.

2) At least for Nikon I know there's supposedly an internal database or 'catalog' of typical image data that it uses for reference purposes, to help it guess at exposure and white balance settings. If that's the case with the Fuji implementation, it could be interpreting the scene based on it's internal reference point, then guessing at some appropriate WB settings. That obviously wouldn't take into account the filter and would return correspondingly warm results.
 
What does AWB mean?

It means one defers decisions about parameter optimization (in this case color temperature) to strangers who can not anticipate every possible circumstance required to optimize the parameter(s) of interest.

Often AWB can not succeed (regardless of camera brand). AWB does not work when a scene is lit by two or more strong light sources with very different color temperatures. A few common examples would be: blue color casts in shadows when a scene is lit by daylight; an interior where one are is lit by tungsten light and another is dominated by outdoor light from windows. AWB is impossible because a single set of color temperature parameters is insufficient.
 
Back
Top Bottom