Why Is There No Reference White Balance

Question

I understand why White Balance is a necessary part of any imaging pipeline, but I don't understand why using a colorchecker on a measured sensor does not have an associated reference equation to consistently create perfectly reproduced colors. I'd imagine that knowing the following variables would fully constrain the problem:

• Quantum efficiency of each color channel in the sensor
• An image of a greycard in the scene lighting
• The rate at which human photoreceptors saturate
• The target display characteristics
• The lighting of the viewing environment

If such a relationship can exist and is known, what is it?

Answer 1

I was more looking for an optimal algorithm to map the colors of a raw image using a known sensor and target display, along with an image of a color card in the scene lighting.

That algorithm is: pick something in your image which you want to render as a neutral gray — a gray card or a gray patch on a color target is ideal — and adjust the multiplier for each of the red, green, and blue channels so that the levels of that target are equal.

That's all there is to it.

There may be an earlier stage which adjusts for the different color response of your camera itself — this is what a camera device profile is for.

And there may be a later stage usually unrelated to anything changed in any specific image where the output is changed to get correct rendering on a given display.

But the white balance part? That's just the simple balancing in the middle.

Answer 2

The usefulness of things such as color checkers is precisely because all of these things are not usually known to the degree needed. Particularly, the final two on your list are often totally unknown at the time the image is taken.

In addition to your list, the exact nature of the lighting in the shooting environment is often not precisely known. This is perhaps the most important reason to use a color checker along with a gray card (which, incidentally, is primarily a tool for calibrating exposure, not color).

Just one example:

You can say outdoors under a sunny sky is a certain "color", say 5500K, but there is still a long list of variables that will affect the exact nature of that light and the results of photos taken under such a sky. These things can affect not only the color along the Blue←→Amber color temperature axis, but also the "tint" along the more or less perpendicular Green←→Magenta axis. A few are:

• The height of the sun on the sky. The closer to the horizon, the more atmosphere the sun's light passes through to reach where you are shooting.
• The atmospheric conditions of the parts of the earth's atmosphere that the sun's light passes through. The amount of water vapor, dust, and other particulates in the air can have a profound effect on what portions of the spectrum of the sun's total output reaches a specific location. This can be constantly changing.
• The amount, density, and angle to the sun of any clouds in the sky. This can be constantly changing.
• The color and reflectivity of the landscape. Sunlight is reflected off everything you see outside during the day. A green lawn will affect the color of light on your subjects differently than a sandy beach, or snow cover, or a field of yellow sunflowers, etc. Foliage and vegetation are highly variable from one type of plant to the next.

The same is true of many environments lit by artificial lighting. Even if the exact type of lighting is known there are still many variables:

• The output of a given type light may not be as consistent as we need. Many types of lighting can vary by significant amounts from one example to the next. The same bulb can vary with the exact amount of voltage flowing through it, will change as the bulb ages, etc.
• Environmental factors play a role here as well. Different floor, wall, and ceiling treatments will affect the color of things in a room.
• It gets even more complex in environments with mixed lighting. The proportion of influence that each source has on a subject will vary by the subjects position in the room, by the camera angle, etc.

In the end, knowing the camera's reaction to specific wavelengths and intensities of light is not enough unless we also know just as precisely the nature of the light falling upon and being reflected upon our subjects.

Answer 3

The basic problem is that the lighting can vary hugely. You can know all there is to know about your sensor and the process after it, but all that does little good if you have no reference for the lighting.

That said, you can have a few limited "canned" lighting situations you can correct for with known profiles. Modern camera sensors are linear, so unlike with film, if you correct for gray anywhere in the picture, it applies to everywhere else subject to the same lighting.

About the only lighting condition that is both recognizable and repeatable enough to be useful is "full sunlight". The means the subject directly illuminated by the sun, no clouds over the sun, and not close to sunrise or sunset. Even then there can be variations due to haze, and what nearby objects light is diffusing from into the shadow areas. However, I've found having a canned "full sunlight" color correction is still useful.

Go out on a nice sunny day within a few hours of noon, and put a bright white but diffuse (not reflective) object on the ground. Take pictures of it somewhat from the side so that the sunlight doesn't reflect off of it directly into the camera. Take a few pictures underexposed from 1 to 3 f-stops.

The best picture is the one with the brightest values but not clipped. Average a section of the picture with only the white object in it. That's your color balance for "white" for future full sun pictures.

Regardless of all that, it's still better to include a white reference in at least one picture of a set taken of the same scene and with the same lighting. Having a available measurement of what white should be is always better than guessing.