Casting Light on Illumination: A Computational Model and Dimensional Analysis of Sources

tion, such as when a thin cloud passes in front of the sun and the illumination becomes more diffuse. Similarly, we We attempt to cast light on illumination by introducing a computational model of light sources and illumination. Spaoften have difficulty guessing the lighting configuration in tially varying illumination is computed using coordinate transa scene, such as in a magazine photograph, even though formations on the manifold of rays in space, and the ray maniwe may feel confident of the surface shapes, materials, and fold is shown to be 4 dimensional. The illumination model can 3-D layout. Based on these observations, it is tempting to be used to recover shape from shading in situations when direct believe that we judge illumination poorly relative to how illumination varies across space and provides the basis for well we judge shape, material, and 3-D layout and that in relating different types of light sources (e.g., an overcast sky, many cases, we cannot judge illumination at all. the sun, a panel light, and a spotlight). A formal definition for Yet, from a computational perspective, it would be puzlight sources emerges that generalizes the standard definition zling if the visual system were less sensitive to illumination of a source as an emitter and leads to a 4-D light source hypervariations than to variations of shape, material, and 3-D cube in which the different types of sources may be embedded.  1997 Academic Press layout. All of these variables are intrinsically coupled in the physics of image formation, and it is impossible to say in general that one variable determines scene appearance