What's the base WB of a camera sensor?

[ntx]

Hi all,
I guess I can't post anything odd as first thread.

I would like to know what is the base WB of a camera sensor. With "base WB" I mean what kind of light I should have on the scene to let the WB algorithm in the camera doing nothing on the RAW data?
I think when the RAW data is processed to adjust the WB, an algorithm (in the camera or in the RAW developer software) reduces the strenght of the light dominant components. Now, doing this will reduce the available dynamic of those components.
On the other side if the WB algorithm amplify the component signals to remove the light dominant, you can end up by clipping a color component.

By intuition I could guess the "base WB" is the sun light. The sun light has all the light frequencies in its spectrum: it's quite linear (no frequency peaks) and continuous.
But when I look at the Bayer matrix data I start having some doubt...

I hope this question is not too odd... I like CiC cause I find many technical explainations about digital photography, so I feel confident about having an answer.

Bye
Jenner

[Paul Coyne]

I understood that if you shot in Raw then the WB was neutral, and that only saving as a jpg would fix it on one setting or another

[pwnage101]

This is a very good question and I've been wondering this myself. Shorter wavelengths of direct sun are filtered out by the atmosphere, only to be introduced from the sides (hence a blue sky) so I think sensors are more likely to be tuned for general daylight rather than the longer, yellowish wavelengths of the direct sun.

Did you know the sun looks white to human eyes outside the earth's atmosphere? Learned that in astrobiology, "the study of aliens" as we called it

[Peter Ryan]

Colour temperature or the colour of light is measured on the Kelvin scale on a range of 0 – 10,000. Try to think of this NOT as a hot/cold temperature but as a colour guide for direct, indirect and manmade light sources.

The Sun, the hottest object we have in our universe generates light at around 5,500 on the Kelvin table. This is a pure white light with no colour cast.

If I asked you “what is the colour a camp fire blaze?” You would answer red.

If I asked you what colour light a fire gave off you often hear people say ‘there was a warm glow around the fire’. That’s right the light has a colour that is a reddish/orange to yellow.

So it is that light has a colour.

The human brain/eye combination automatically adjusts to the colour of light and no matter the light source if you are reading a book the pages will always appear white.

Your camera’s white balance meter is designed to try an emulate this and bring all subjects, no matter what light source, to the comparable colour as if they were illuminated by what white light.

The Auto White Balance does a pretty good job where there is one light source but struggles with mixed light sources.

When shooting in jpeg the camera processes the signal as read by the camera. When shooting in RAW the camera pegs the white balance (akin to say the base slider on a music amplifier) but it is not set and can be changed when you process the image in your RAW converter.

If you want to have the camera set the white balance at the same point on the Kelvin scale as seen at the time of shooting then you would need to have a camera where you can manually adjust the Kelvin scale or approximate it using the White Balance modes and fine tune these using the MIR scale (the intermediate values on the Kelvin scale for wont of easy explanation) adjustment available on many DLSR.

My understanding is the white balance adjustment doesn’t reduce the light in any particular range and hence luminosity of the image but it does change the colour of the light to reflect any adjustments you make along the Kelvin scale.

I am sure others will respond to this thread so keep watching.

[Colin Southern]

Hi Jenner,

I can't put a figure on it, but "for a spectrally neutral target, whatever temperature give equal output across all 3 channels" -- so if you wanted to experiment, shoot a grey card and when the R, G, and B histograms are all aligned, you'll have your temperature. I'd suspect 5500 kelvin would probably be pretty close.

To comment on your original question, yes, it's certainly possible for a dominant colour to blow a particular channel - quite common with bright red flowers.

In terms of adjusting the colour temperature of a RAW capture though, it's surprisingly "the closest thing to a free lunch" that you're likely to get; a component of the RAW conversion process is a thing called a "decoder ring" which refines the actual shades of red, green, and blue filters used in the bayer array. When one adjusts the colour temp in ACR what they do is redefine these colourimetric interpretation values, not scale the axis, so the conversion is pretty much lessless in terms of dynamic range (but of course doesn't help if the channel is blown to start with).

At the end of the day it's somewhat of a moot point though - sure, we can be careful of our exposures when photographing predominantly bright objects, but apart from that, I'll be damned if I'm going to wait for a "colour temp neutral day" before I shoot just to maximise dynamic range!

[ntx]

Thank you all for your replies.

As Colin said I'm not going to go out taking some photos only when the light is neutral. But suppose I'm about to code a program that adjusts the WB of a given Bayer matrix and I don't want to lose any color gradiation... the information about the base WB could be used to adjust the data in an adequate number of bit instead of applying the "standard" multipliers used to adjust the WB (if you take a look at how DCRaw adjust the WB all becomes clearer).

To get the base WB I can't use the camera histograms: they are made after adjusting the WB... so, on a grey card, they are always aligned (if I don't use the wrong settings, like tungsten on sun light or viceversa).

I think I have to do some test. I could take a RAW shot of something white (snow, paper) under sun light (using the correct exposition). After that I could take a look at the data to see if I have some color dominant.
I have just tried to do something in this way but the shot was took when the sun was about one hour before sunset... the result gave me a cyan dominant in the shot: may be 5500 K is the correct answer for my question?

Bye
Jenner

[Colin Southern]

I think I have to do some test. I could take a RAW shot of something white (snow, paper) under sun light (using the correct exposition). After that I could take a look at the data to see if I have some color dominant.

You really need to use a spectrally neutral card that won't suffer from metamerism. As an example, "white" paper may well have UV brighteners that will reflect more energy in the blue/violet space of the spectrum.

Have a read up on UniWB, I think it might give you more info that you're seeking.

One of our members - _GUI_ wrote up this article about it ... http://www.guillermoluijk.com/tutori...b/index_en.htm

[ntx]

Colin, that article contains the answer to my question. The UniWB is exactly what I look for: with this kind of WB all the multipliers for adjusting it are set to 1.0 ... that means the WB isn't applied.

Thank you for sharing this article!

Bye
Jenner

[Colin Southern]

Colin, that article contains the answer to my question. The UniWB is exactly what I look for: with this kind of WB all the multipliers for adjusting it are set to 1.0 ... that means the WB isn't applied.

Thank you for sharing this article!

No worries ... I'm sure Guillermo wouldn't mind if you sent him a PM if you'd like to discuss it further.