Color Constancy is a feature of the human color perception, in which the perceived color of an object remains relatively constant. During the day the ambient light changes, so the illumination conditions change and still, red strawberries would still be perceived as red strawberries. Color Constancy is the ability of human vision to perceive and recognize the color of an object independent from the illuminance. The visual system, the eyes and the brain calculate the average illumination of a scene and then subtract those conditions, so that the colors remain relatively constant. This is why a blue object looks blue, whether it is looked at under the midday sun or a dim sunset.
Abbildung in dieser Leseprobe nicht enthalten
Figure 1: Even though this image appears to contain a lot of different colors, the image is only made out of red. Source: Internet, https://www.youtube.com/watch ?v=aHtjhCxRla4
Wandeil (1995) wrote, that “[...] color appearance is a mental explanation of why an object causes relatively more absorptions in one cone type than another object. The physical attribute of an object that describes how well the object reflects light at different wavelengths is called the object’s surface reflectance.”
The work of Maloney & Wandell ( 1985) is based on different theories of other scientists. For example, Land’s retinex (retina and cortex) theory (1977), that explains the Land Effect (1971), by assuming that both, the eye and the brain, are involved in the process. This experiment involved a display, which was illuminated with three white lights. Those were projected through three filters (red, green and yellow). First, the test person has to adjust the lights, so that the patch is appearing white. The intensities of the filters were measured then by the experimenter while asking the test person to identify the color of the neighboring patch. Color Constancy is shown, when the new patch appears the same color, even when the experimenter adjusts the lights the same way, the test person adjusted the lights for the white patch. This and other retinex algorithms were developed over time. Land and McCann also developed a computer program to imitate the retinex processes in human physiology (Land & McCann, 1971).
Buchsbaum (1980) formulated a comprehensive mathematical model to account for color constancy. Our visual system is able to recognize true object color under various spectral compositions (Color Constancy). Because of that, it is assumed that the visual system estimates the illuminant. Buchsbaum (1980) postulated that this estimate of the illuminant is made on the basis of spatial infomiation from the entire visual field, which is then used by the visual system to get an estimate of the reflectance of the object. He managed to compute color descriptors that are independent of the ambient light in an image, but only if the average spectral reflectance of the objects in this image are known. To compute the color descriptors where the average spectral reflections of the objects in an image are unknown, the authors (Maloney & Wandell, 1985) suggested to improve Buchsbaum’s result. The idea of the paper was to recover the surface spectral reflectance from an image where the average spectral reflectance was unknown. This intent should be accomplished by developing an algorithm, which an image-processing system can use to assign colors.
To understand the algorithm that was developed in the paper (Maloney & Wande 11, 1985), a few preliminary mathematical definitions have to be introduced. In their work, Maloney and Wandell used a visual sensing device which was analogous to a retina. That system contained a lens that focused the light from a scene onto a planar array of sensors.
The reflection of light by a matte surface can be described by a simple mathematical formula: c(A) = s (A) e (A). In this formula, the illuminant spectral power distribution is described as e(A) and the body reflectance is described as s(A) (Wandell, 1995). In foundations of vision Wandell (1995) describes the changes of the reflected light from objects as the illuminant change also. In the following image two different light sources ((a) tungsten bulb and (b) sunlight at a blue sky) are used. The shaded panels on the left shows the spectral power distributions of the two different light sources. The three graphs on the right shows the reflected light from a red, green and yellow paper, when they are illuminant by this source. The bar plots above these graphs shows the three cone absorption rates caused by the color signal:
Abbildung in dieser Leseprobe nicht enthalten