Universal White Balance

I'm trying to use the UniWB technique to get a closest preview of the RAW histogram from the RGB histogram and I'd like some advices from gurus who master well this technique. Here are the camera parameters :

1. I have set the style to neutral, contrast 0, saturation 0, sharpness 0, tint 0 : is this correct ?

2. I have set the color space to Adobe. Is this better than setting it to sRGB ?

3. is it better to display the RGB histogram or the luminance histogram ?

4. Can someone give me a link to download a uniwb file for the 5D classic camera ?

Thanks.

It is very easy to make a Uni-WB custom WB setting. (BTW, the "Uni" is for "unity", not "universal", because it is a WB setting in which all the channel multipliers are equal to one.) When we make a custom WB setup we show the camera a photo of a target that should be neutral grey (R=G=B) but isn't because of the light and the camera figures out how much the red and blue channels need to be multiplied in order to make it so. So we fool the camera by showing it a photo in which R,G and B are equal and the camera figures that all the multipliers should be 1.0. There are two cases in which this happens - a.) when the photo is totally blown out, grossly overexposed, so that all three channel values are 255, and b.) when it is totally underexposed, clipped to black. So you can make a way overexposed photo and use that to make a custom WB or you can take a shot with the lens cap on and use that. Theoretically the white photo method should be better because you can never get a purely black photo, there is always noise, but in practice the black frame shot works well enough and is easier to do. Just put ISO down to 100 and f-stop and shutter speed all the way up.

Thanks Elie.
I have found a file here : http://www.guillermolu​ijk.com/tutorial/uniwb​/index_en.htm
( See at the end of the page, BTW there are also files there for G9,300D,350D,400D,B450​D,500D,550D,20D,30D,40​D,50D5D,5DMkII,7D,1Ds,​1DMkII,D60,D90,D300,D7​00,D800)

I have tried the lens cap method but it seems to not work very well.

I have done 2 test shots : one with WB set with the file provided above, and another one with WB set with my file (lens cap).

With exiftool I get the following values :
first file : 1015 1024 1024 986
my file : 679 1024 1024 512

I guess the perfect values are 1024 1024 1024 1024, so the first file is better.
If someone has a better file than this one, please give me a link.

Interesting, with the black frame on a 5D2 I get 1024/1024/1024/1023.

1024/1024/1024/1023 ! wow ! That's almost perfect !

May be you can send your file to <gluijk at hotmail.com> as he was asking for better files...
"Anyone trying the procedure and achieving multipliers close to 1.0 (let us say with a deviation lower than 10%), or manages to improve the previous deviations, please send them to me to add the files to this list."

I don't have the file; I never downloaded it to the computer and that card has been formated a lot of times since then.

I'd just use the lens cap approach. I figure that it's good enough for any but the most challenging color scenarios.

But I honestly never use UniWhiBal anyway...!

Quick question (Canon S95), re.: the values of the WB multipliers (why one of them is not set to 1, and what does it mean to the in-camera histograms - are they scaled down?)

Example:
The Daylight WB multipliers: 1541 870 870 1840
In another shot the same preset Daylight WB multipliers read: 2001 1006 1006 1818
UniWB as set by oversaturating pixels read (at least in one image): 1280 1280 1280 1280

Why the coefficients are scaled so that none is set to 1000?
Why preset coefficient change from shot to shot?
How the UniWB coefficients affect interpreting the in-camera histogram: they are equal, but greater than one - does it mean that the in-camera histogram will show clipped histograms even if no RAW data is clipped?

Cheers!

In Canon EXIF the numbers are scaled to G = 1024 which is 1.024, very close to 1.0.

No idea. Where are you getting these numbers from?

Yes, I think it would.

Thanks Eli,

If the values were 1024, I wouldn't complain, but look at the values posted in the original post: they are far from 1.0! Why?

The values are from makers EXIF, in CR2 files taken on Canon S95. Even the preset WB's values change from shot to shot.

If UniWB values are 1280 all and that's the actual multipliers that the camera uses for histograms, that defeats the whole idea of using UniWB, doesn't it?

I completely don't understand why the preset values vary from shot to shot. I have never seen that in any of my cameras (350D, 40D, 5D2) and the only thing I can think of is that in the S95 Canon is combining some auto exposure adjustment with the WB,but I can't see any advantage in doing so instead of doing it in two separate operations. And yes, a Uni-WB that is not close to 1 is not really a Unity White Balance.

The advice I usually give to anybody who doesn't want to mess with Uni-WB but wants good ETTR is to look only at the green channel histogram, because in most light conditions the green channel should be unaffected by WB and if the green is not clipped the chances are very high that the other channels won't be clipped. However, considering what you have discovered about the S95, I really don't know whether the green histogram can be trusted.

You can download Raw Digger and do a series of tests to see at what point the camera lcd indicates clipping although Raw Digger shows that the Raw is not clipped. At least you will know what to expect from the camera and be able to make the appropriate allowances.

As a side note - if you use a camera that is supported by the newest version of Magic Lantern, there is an option to set the WB according to WB multipliers for each channel - that is, instead of shooting a dark or completely blown image and assigning it as your custom WB image, you can explicitly define R, G1, B, G2 = 1 in the Magic Lantern menu item for WB. Pretty cool.

kirk