|
| Author | Thread |
|
|
11/19/2007 04:12:41 PM · #1 |
There's a thread on "Auto White balance is an Abomination" that is getting pretty long and opinionated. It was suggested this might be a good post to put somewhere else so it doesn't get lost in the chaff, so here it is for what it's worth:
****************
Let's try to explain how AWB (auto white balance) works, OK?
Presumably we all are aware that light meters base their calculated exposure on the assumption that the thing being metered is zone 5, a middle gray. This means if you meter a white wall and use the recommended exposure you get a gray, underexposed wall. And if you meter a black wall and use the recommended exposure, you get a gray, overexposed wall. When your camera is using averaging meter modes that meter the whole scene, it assumes that all the values, if mixed together, would come out to an average gray.
So if you paint a wall half white, half black and frame the image so you see 25% white and 75% black, then using the averaging meter mode you will get an overexposed image. And if you reverse that (75% white and 25% black), you will get an underexposed image.
Now, Auto White Balance works the same way, except with colors. AWB is basically assuming equal amounts of all colors in the image, so it should average out to a neutral balance. If you shoot a scene that is primarily sky, a brilliant blue sky, AWB will try to warm up that sky and it will look greenish in proof. If you shoot a scene that is primarily bright, yellow fall foliage, AWB will bring in too much blue and it will look dull in proof.
In-camera WB settings are, in fact, based on "kelvin units" of color temperature. Depending on manufacturer, "daylight" setting will be around 5500 kelvin. "Tungsten" setting will be around 3200 kelvin, the approximate color temperature of fresh, tungsten-filament bulbs (they get warmer as they get older). So with a camera that allows you to set WB in kelvin units, there will be no difference between a 3200 kelvin WB and a tungsten white balance; the "named" settings are just commonly-used WB settings that are preset so you can access them without taking the extra steps to manually dial up the correct setting in kelvin units.
Now you've all had the experience, I'm sure, of sitting inside by a window, under room lighting, and looking up and seeing a beautiful, blue-magenta sky at around twilight, right? The room looks "normal" (the in-brain AWB has neutralized the tungsten) and the sky colors are really rich and striking.
So you go outside to admire the sky, and in a matter of moments the sky seems to "desaturate" and lose its impact; but if you look INTO the house through the window, the interior seems very yellow. That's the brain's AWB resetting to the ambient color temperature of the exterior, which is MUCH cooler then the interior, tungsten lighting.
As photographers, we have to make our decisions as to where we want the balance to be between these extremes. Inside, looking out, tungsten WB is a better choice usually; the visible room is correctly rendered, and the scene through the window is more blue and saturated. Shooting from the outside in, the reverse is true; if we balance for the interior, the outside walls of the house will look unnaturally blue.
Many times, the "correct" WB setting will be between the two extremes. In theory, this is what AWB does for you; it strikes a balance between the two extremes. But AWB can be fooled if it is the subject itself that is anomalously colored, not the light that is falling on the subject. And AWB can give disappointing results when it is the light itself that is colored and that aspect of the light is what you are trying to capture. So, as a rule, if time permits it is better to use a specific WB setting than it is to use AWB, because then you control the rendering of the scene based on your learned response to these variables.
R.
edit to remove reference to "degrees" kelvin
Message edited by author 2007-11-21 03:13:54.
|
|
|
|
11/19/2007 04:20:54 PM · #2 |
|
|
|
11/19/2007 04:56:43 PM · #3 |
Thanks, though I think my head hurts from the whole thing. (It's still easier than calculating exposure parameters in my head)
I find I am using auto white balance as the default setting for taking quick shots, but I frequently select one of the presets depending on what looks right. Since my camera does not do raw, I suppose it matters a little more for me. |
|
|
|
11/19/2007 05:28:59 PM · #4 |
Originally posted by yospiff:
Since my camera does not do raw, I suppose it matters a little more for me. |
VERY much so. Especially when shooting "mood" shots, the kind where you look at the light and say "OMG! That's lovely!" and snap off a shot, and then are bitterly disappointed at how blah it looks; that's the AWB sucking all the magic out of the light...
R.
|
|
|
|
11/19/2007 07:11:20 PM · #5 |
That's a great explanation of color temperature, the most important dimension of white balance. It sets the color of the light on a scale from red to blue.
But there's another dimension that also plays a part: the amount of green. Natural light is green-neutral, as is tungsten light. But fluorescent light has a lot more green in it. The brain adjusts for it automatically just as it does other colors of light, but the camera (or RAW converter) needs to know how green the light is to properly render the colors.
In RAW converters, this is generally called "Tint", and there's a slider for it right by the color temperature slider. I haven't seen a camera that lets you set it directly, but a preset for fluorescent tubes will have a high tint value. And Custom WB will adjust for the amount of green in the reference shot. So will AWB of course, and that can throw off your color for subjects that have an abnormally high or low amount of green in them. |
|
|
|
11/21/2007 03:02:59 AM · #6 |
Originally posted by Bear_Music:
In-camera WB settings are, in fact, based on "degrees kelvin" of color temperature. Depending on manufacturer, "daylight" setting will be around 5500 degrees kelvin. "Tungsten" setting will be around 3200 degrees kelvin, [ . . . ]
|
Personal pet peeve, ever since I learned about the Kelvin scale in high school, so please forgive my nit-pickiness... I just keep seeing this more and more these days (in photography most of all, but in other sciences which use kelvin, too).
By convention since around 1970, there is no such thing as "degrees kelvin." It's just "kelvins" -- "5500 kelvins." It can be abbreviated "K" (again, no degrees -- just "K" -- "5500 K").
Who says so? The International Bureau of Weights and Measures, the General Conference on Weights and Measures and pretty much every other "standardization" body you can name who has an interest in measuring things. :P
If you want to show uber-geekiness (and who doesn't? -- okay, maybe I'm just weird...), since kelvin in an SI unit, all the usual SI multiples apply... So, 5500 K == 5.5 kK == 5.5 kilokelvins. How cool is that? Err, warm... whatever.
Sorry, it's late, I'm stuck on a conf. call for work, and I couldn't stop myself from rambling about this. Feel free to ignore me; I just had to get that off my chest... I know, it's a losing battle -- I'm sure "degrees kelvin" has probably made it way into camera documentation and all kinds of stuff; it will probably end up so commonly "mis-used" that a standards committee with change it so either form is correct, but until then my pet peeve lives on... ;)
|
|
|
|
11/21/2007 03:09:49 AM · #7 |
Originally posted by cdrice:
I'm sure "degrees kelvin" has probably made it way into camera documentation and all kinds of stuff; it will probably end up so commonly "mis-used" that a standards committee with change it so either form is correct, but until then my pet peeve lives on... ;) |
Actually, I'm not sure but I think that may be backwards. I remember way back when (60s) "degrees kelvin" was standard in photography; or at least that's how I remember it. And I do remember being educated otherwise, to "5500 kelvin" at some point, but I guess I have lapsed.
You are, of course, correct; the word "degrees" is extraneous here.
R.
|
|
|
|
11/21/2007 03:27:32 AM · #8 |
| Thank you for this! I know this will help me a lot. |
|
|
|
12/23/2007 08:59:31 PM · #9 |
Bear - thanks for referencing this again in another thread. I had read it a month ago and was glad for the reinforcement.
|
|
|
|
12/23/2007 09:39:02 PM · #10 |
Originally posted by cdrice:
Originally posted by Bear_Music:
In-camera WB settings are, in fact, based on "degrees kelvin" of color temperature. Depending on manufacturer, "daylight" setting will be around 5500 degrees kelvin. "Tungsten" setting will be around 3200 degrees kelvin, [ . . . ]
|
Personal pet peeve, ever since I learned about the Kelvin scale in high school, so please forgive my nit-pickiness... I just keep seeing this more and more these days (in photography most of all, but in other sciences which use kelvin, too).
By convention since around 1970, there is no such thing as "degrees kelvin." It's just "kelvins" -- "5500 kelvins." It can be abbreviated "K" (again, no degrees -- just "K" -- "5500 K").
Who says so? The International Bureau of Weights and Measures, the General Conference on Weights and Measures and pretty much every other "standardization" body you can name who has an interest in measuring things. :P
If you want to show uber-geekiness (and who doesn't? -- okay, maybe I'm just weird...), since kelvin in an SI unit, all the usual SI multiples apply... So, 5500 K == 5.5 kK == 5.5 kilokelvins. How cool is that? Err, warm... whatever.
Sorry, it's late, I'm stuck on a conf. call for work, and I couldn't stop myself from rambling about this. Feel free to ignore me; I just had to get that off my chest... I know, it's a losing battle -- I'm sure "degrees kelvin" has probably made it way into camera documentation and all kinds of stuff; it will probably end up so commonly "mis-used" that a standards committee with change it so either form is correct, but until then my pet peeve lives on... ;) |
[off topic - sorry]
The whole degrees kelvin or just kelvin thing is actually starting to change back to "degrees kelvin" and "kelvin degrees" (different things) same with rankine. Or at least things are in all my engineering and thermodynamics classes.
And 5500 kilokelvin is just a mild 5227 C. :)
[/back on topic]
Great write up, it was a great read. Thanks for doing this!
Message edited by author 2007-12-23 21:42:45.
|
|
|
|
Current Server Time: 04/24/2024 01:23:00 AM  |
Home -
Challenges -
Community -
League -
Photos -
Cameras -
Lenses -
Learn -
Prints! -
Help -
Terms of Use -
Privacy -
Top ^
DPChallenge, and website content and design, Copyright © 2001-2024 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 04/24/2024 01:23:00 AM EDT.
|
