Vincent Jolivet, Dimitri Plemenos and Mateu Sbert

Laboratoire MSI, Université de Limoges,

1, rue d’Isle,F-87000 Limoges, France.


homepage URL:,

Institut d’Informŕtica i Aplicacions, Universitat de Girona

Lluis Santaló s/n, E-17071 Girona, Spain.




In this paper a complete presentation of Pyramidal Hemisphere subdivision method is given. This method improves Monte Carlo radiosity by sending more rays towards selected directions. More precisely, we determine regions of the scene where the distribution of the power must be done more accurately. The number of rays sent in a direction is function of a pyramidal region, defined by the center of the shooting patch and a spherical triangle on the surface of a hemisphere surrounding the patch, and the number of patches contained in this region. Thus, the rays shot from a patch have not all the same power. This method allows us not only to obtain fine details much sooner and with lower cost, but also the overall efficiency is considerably increased. Next we present a new way in order to calculate the number of rays sent in a region by using a dynamic estimation of the visible patches in this region.

Keywords: Monte Carlo radiosity, hemisphere subdivision, heuristic search.