Is "White Balance" relative or absolute?

[pnodrog]

Either the wiring in your house was rubbish or the delivering of the power. I hardly can believe that last in a city as Ottawa.

The point here is that when one wants to calculate with the color temperature or whatever else one has to use figures. A programmer can't work with his feelings.

George

You should acquaint yourself with power distribution systems, auto stepping transformers, load variations and voltage tolerance etc.

Cameras only estimate WB and like the estimations they do for exposure will have errors when presented with unusual lighting conditions. It makes little difference how good programmers or their programs are if the initial value is incorrect.

A camera is not a device for measuring colour temperature or WB it merely takes what could be best described as an educated guess. No amount of precision in subsequent calculations will correct an incorrect initial value.

[Manfred M]

Either the wiring in your house was rubbish or the delivering of the power. I hardly can believe that last in a city as Ottawa.

The point here is that when one wants to calculate with the color temperature or whatever else one has to use figures. A programmer can't work with his feelings.

George

George - under varying load conditions both the frequency and line voltage are affected. That's just how power distribution systems work. I believe that in North America; the electrical grid will deliver line voltage from 110V - 120V. I believe much of the rest of the word runs from 220V - 240V. Change in line voltage will affect the colour output of a tungsten bulb used in an enlarger. That's also why dichroic head colour enlargers use a regulated power supply.

While a programmer does not work with feelings; he or she will work on assumptions. These tend to work well under many, but not all conditions. When I look at my camera manual, Auto White Balance (AWB) only works for colour temperatures between 3500K and 8000K. That means the programmers have written the camera firmware in such a way that the white balance will be incorrect under normal tungsten light (3000K) and when the sun is close to the horizon (2500K). It will also not give accurate readings when I point my camera towards the north on those nice sunny days where colour temperatures are in the 10000K - 20000K range.

However, I can manually set the colour temperature from 2500K - 10000K, so there is obviously a gap in what the camera is capable of measuring and the settings I can choose.

[Manfred M]

I'm hoping that most people here mean "correlated color temperature" (CCT), not black body temperature. I'm also hoping that most people are not thinking that two lamps of the same CCT have the same chromaticity (color):

A Black Body radiator is a theoretical construct, not a "real world" light source. Some "real world emiiters" do emulate what a Black Body source would act like fairly well through part of their operating range; i.e. linear response. This is how devices like an optical pyrometer work.

Materials that emit light when heated should have reasonably consistent spectral output over part of the range, but light sources that use other means to emit photons (fluorescent or LED lights) will definitely have a different spectral response.

[xpatUSA]

George - under varying load conditions both the frequency and line voltage are affected.

That would not be frequency, Manfred. The entire National Grid runs at a constant 60Hz.

[dje]

Based on this evidence I'm thinking that while a value for temperature is being deduced based on the raw data the method for doing this is differs from one raw processor to another and I must conclude is not precisely defined.

Yes David I believe that is the case. Also don't forget that the WB settings the camera comes up and puts in the raw file (As Shot), are only estimates to start with. If you have a shot with a neutral gray card in the frame, you can use this to get an accurate WB adjustment in the raw processor using the white point selector.

Another thought I've had is that conceptually WB is something that only applies to raw processing and that I'm guilty of misusing the software when trying to apply the tool to previously developed images. I suspect that other color management tools could be used to achieve the same result but that it may not be as simple as using the WB slider.

Finally, do experienced photographers who know what they are doing use the WB tool on previously developed images?

Well I dont usually use a WB adjustment on a jpeg. You can do some fairly controlled color adjustments with an HSV or HSL adjustment (on raw or jpeg). This allows you to alter the hue and saturation of individual color ranges (eg red, yellow, green blue etc).

Dave

[Manfred M]

Finally, do experienced photographers who know what they are doing use the WB tool on previously developed images?

The way I was taught to handle this in an advanced college level course on colour correction was with curves using Photoshop. The issues is that the colour balance issues in a JPEG are usually non-linear. It also requires the use of the Info window so that the individual colour readings of the sample point can be read.

Step 1 - identify parts of the image that are likely "white" and "black". From a practical standpoint what one is trying to look for is an area that is effectively neutral and close to white, without having any of the colour channels clipped and something that appears to be close to black, again with enough colour data to ensure that clipping does not occur.

Step 2 - Once the appropriate points have been selected. In the attached image I show a "white point" (240,243,246) and a "black" point of (12,16,19). These values are the R, G and B readings. in both cases the green channel has values that lie between the red and blue channels. This tells me that I need to red values up and the blue values down to get the points to be a neutral tone. If I don't get points that give me this type of relationship; where one of the channels has values in the middle, I have not picked neutral points for my measurements and I need to test a different black sample point and /or white sample point.

Step 3- make the adjustments individually on the red and blue channels by moving them vertically, so that we get a neutral reading for all three channels. If you look at the second numeric value (to the right side of the "/"), I get "white" readings of (240,241,241) and "black" readings of (14,15,14). Sometimes I can get readings that are 100% neutral and other times I might be off by a single digit in one of the values, which is still very good.



Unfortunately, I can only show one channel at a time in the screen capture, which is why only the red channel with adjustments is shown. Curves are the single most powerful tool in editing work, but are not trivial to work with. I hope that this helps.