Efficient inverse lighting: A statistical approach

Abstract

Given a scene to illuminate satisfying a specific set of lighting intentions, the inverse lighting techniques allow us to obtain the unknown light source parameters, such as light position or flux emission. This paper introduces a new inverse lighting approach that uses the radiosity mean and variance to define lighting intentions of a scene. It is shown that these statistical parameters can be obtained without the previous calculation of the radiosity of the scene. Avoiding the explicit computation of the illumination of the scene results in a drastic reduction of the time required for the inverse process. This approach also provides a methodology that transforms a current set of lighting intentions into a single lighting intention with statistical parameters. The tests show that the processing time for solving the inverse problem can be reduced to a few seconds in most cases, improving previous work