Design Article
Image sensor color calibration using Zynq-7000 SoC
Gabor Szedo, Steve Elzinga, Greg Jewett, Xilinx Inc.
11/14/2012 4:15 PM EST
White balance
You may view any object under various lighting conditions—for example, illuminated by natural sunlight, the light of a fire, fluorescent or incandescent bulbs. In all of these situations, human vision perceives the object as having the same color, a phenomenon called “chromatic adaptation” or “color constancy.” However, a camera with no adjustment or automatic compensation for illuminants may register the color as varying. When a camera corrects for this situation, it is referred to as white-balance correction.
According to the top equation at the right of Figure 1, describing spectra of the illuminants, the reflective properties of objects in a scene and the spectral sensitivity of the detector all contribute to the resulting color measurement. Therefore, even with the same detectors, measurement results will mix information from innate object colors and the spectrum of the illuminant. White balancing, or the separation of innate reflective properties R(λ) from the spectrum of the illuminant I(λ), is possible only if:
• Some heuristics, e.g. the spatial frequency limits on the illuminant, or object colors are known a priori. For example, when photographing a scene with natural sunlight, it is expected that the spectral properties of the illuminant will remain constant over the entire image. Conversely, when an image is projected onto a white screen, spectral properties of the illuminant change dramatically from pixel to pixel, while the reflective properties of the scene (the canvas) remain constant. When both illuminant and reflective properties change abruptly, it is very difficult to isolate the scene’s objects and illuminants.
• Detector sensitivity S(λ) and the illuminant spectrum I(λ) do not have zeros in the range of spectrum observed. You cannot gain any information about the reflective properties of objects outside the illuminant spectrum. For example, when a scene is illuminated by a monochromatic red source, a blue object will look just as black as a green one.
Next: Prior methods
You may view any object under various lighting conditions—for example, illuminated by natural sunlight, the light of a fire, fluorescent or incandescent bulbs. In all of these situations, human vision perceives the object as having the same color, a phenomenon called “chromatic adaptation” or “color constancy.” However, a camera with no adjustment or automatic compensation for illuminants may register the color as varying. When a camera corrects for this situation, it is referred to as white-balance correction.
According to the top equation at the right of Figure 1, describing spectra of the illuminants, the reflective properties of objects in a scene and the spectral sensitivity of the detector all contribute to the resulting color measurement. Therefore, even with the same detectors, measurement results will mix information from innate object colors and the spectrum of the illuminant. White balancing, or the separation of innate reflective properties R(λ) from the spectrum of the illuminant I(λ), is possible only if:
• Some heuristics, e.g. the spatial frequency limits on the illuminant, or object colors are known a priori. For example, when photographing a scene with natural sunlight, it is expected that the spectral properties of the illuminant will remain constant over the entire image. Conversely, when an image is projected onto a white screen, spectral properties of the illuminant change dramatically from pixel to pixel, while the reflective properties of the scene (the canvas) remain constant. When both illuminant and reflective properties change abruptly, it is very difficult to isolate the scene’s objects and illuminants.
• Detector sensitivity S(λ) and the illuminant spectrum I(λ) do not have zeros in the range of spectrum observed. You cannot gain any information about the reflective properties of objects outside the illuminant spectrum. For example, when a scene is illuminated by a monochromatic red source, a blue object will look just as black as a green one.
Figure 1 – Spectral responses of the “standard observer”
Next: Prior methods
|
|
|
|
|
Navigate to related information