In the previous example I discussed the process of determining the colors of objects under white or neutral lighting conditions. In principle the hue of the color that a camera detects while filming or photographing, is dependent on the color temperature of the light source used. To complicate matters further, our brain automatically white balances the images we see with our eyes. As such, we are able to distuingish colors under different lighting conditions. To get rid of the inevitable color casts that the camera records, there are a large number of white balance settings. Let’s delve a little deeper into this subject. Take the following example:
Here are two photographs of the same scene taken with different white balance settings. There are two light sources. One is daylight in the background, and the other is Tungsten light in the foreground. Daylight approaches neutral lighting conditions, while Tungsten light is much warmer. By setting the camera white balance to 5600K we assume daylight conditions, and the man has the orange hue of Tungsten light. Conversly, if we set the camera white balance to 3200K, we compensate for the orange hue of the Tungsten light, giving the man daylight skin tones, but the background colors become blue.
As I’ve stated before, white or neutral lighting produces a unique set of colors, so in principle, if we apply a color balancing algorithm to a set of photographs with different white balance settings, the algorithm should produce the same colors for both photographs. First we apply the automated color restoration/balancing algorithm to the first photograph, where we balance the colors using the background only (the algorithm assumes a single light source):
As expected the outcome is very similar to the photograph we started with, as the background represents daylight conditions, which is very close to neutral or white light conditions. Next we do the same for the second photo:
As you can see, the latter result is practically identical to the first result, so the test is successful. I can now say with confidence, that the color balancing algorithm produces the correct colors under neutral or white lighting conditions. Manually adjusting the curves of the red, green, and blue color channels can provide you with an infinite set of color combinations, where a large number of settings may produce colors that seem correct. However, there is only one setting for each color intensity that produces the actual colors of the photographed or filmed objects (under neutral or white lighting conditions). That’s what the algorithm can reproduce.