This page is not fully translated, yet. Please help completing the translation.
(remove this paragraph once the translation is finished)
1. Color Constancy
The white balance of a camera originates from the color constancy of human vision, which is the mechanism by which the human visual system maintains the perceived color of objects under illumination from light sources with different spectra. As shown in the figure, as the color temperature of the light source changes from warm to cool light, the human eye consistently perceives white clothing as a neutral color; this phenomenon is known as color constancy.
2. Camera Sensor Output Under Different Light Sources
Camera sensors, however, lack this characteristic and merely passively record physical light signals. Taking an RGB sensor as an example, this causes the RGB values of white clothing in the image to shift as the light source changes.
3. Image Reference White Point
When we view an image on a display, since the display is a self-luminous device and is independent of the surrounding lighting environment, the human visual system cannot obtain the necessary contextual information, causing the color constancy mechanism to fail.
Therefore, a display requires a fixed reference white. To make the display show this white or other neutral colors like gray, the condition R=G=B must be satisfied.
When we use a display to directly view the images output by the camera sensor under the three aforementioned lighting conditions, the clothing appears in three different colors on the display due to the changes in the corresponding RGB values. In contrast, the clothing observed at the actual shooting scene, unaffected by the lighting environment, appears almost always white.
4. Auto White Balance — Equipping Cameras with a Color Constancy Mechanism
Given that image sensors lack a color constancy mechanism similar to that of the human eye, color digital cameras are often equipped with an auto white balance function. This function uses algorithms to automatically adjust the ratios among the pixel values of the three RGB channels in the image, making the RGB values corresponding to neutral colors in the scene equal. This ensures that they are still reproduced as neutral colors on the display, thereby simulating the color constancy of human vision.