Algorithms compensate for color-camera inaccuracy
ANDREW WILSON
The discrepancy between what the eye sees and the images that cameras acquire becomes obvious when lighting conditions or illumination conditions change. This is because human vision automatically performs computations on the retinal image to compensate for such factors. According to Zia-ur Rahman, vice president of R&D at TruView Imaging (Hampton, VA), these computations produce a high degree of dynamic range compression and color consistency, whereas camera systems merely record the images.
Rahman and his colleagues are now developing algorithms that approximate the way the human retina processes images. They are at work defining algorithms, dubbed Retinex algorithms, that automatically compute images that combine dynamic range compression and color consistency with correct lightness and color rendition. "Ultimately," says Rahman, "this should produce processed images that resemble the human observation of the scene."
Already, the process has found use. The Isothermal Dendritic Growth Experiment that flew aboard the space shuttle Columbia last March was designed to photograph dendrites as they grow from the molten state while being supercooled orbiting the Earth. By processing the image using techniques developed by Rahman and his colleagues, researchers can better understand dendritic growth.
