Authors: [tex2html_wrap4120]M. Langer, S.W. Zucker
Investigator username: zucker
Subcategory: computer vision
Diffuse light sources, such as the sky on a cloudy day, reveal different geometric information about a scene than point sources. According to the standard model for point sources, luminance depends primarily on the surface normal. Under diffuse lighting conditions, however, surface luminance depends primarily on the amount of sky visible from each surface element, with surface normal and mutual illumination of secondary importance. We model this dependence in terms of a surface aperture function, and study its main properties. From a Dominating Sky Principle, an approximately linear constraint between image intensity and surface aperture emerges. We show how such a constraint can be used to recover a depth map from an image. A curious difference from the classical problem is uncovered. When one assumes a point light source, the local geometric constraints of the Shape from Shading problem lie along the surface. However, in the diffuse lighting problem, the local geometric constraints are found in the ambient light above the surface. Such constraints also provide an insight into understanding images whose perception violates the classical image irradiance model.