 
Summary: Accurately blending the occlusion values
and directions between two cones
Supplemental material to
Fast Precomputed Ambient Occlusion
Mattias Malmer1, Fredrik Malmer1, Ulf Assarsson2,3 & Nicolas Holzschuch3
1 Syndicate Ent.
2 Chalmers University of Technology
3 ARTISGRAVIR/IMAG INRIA
July 10, 2006
1 Introduction
In this supplemental material, we give a more accurate algorithm for blending the occlusion values between
two ambient occlusion cones. We assume that we have several occluders in a scene. For each occluder,
we can compute the ambient occlusion as a cone of occluded directions, defined by a direction d and an
aperture . Combining the occlusion between two occluders implies computing the direction and aperture
of the equivalent cone.
Ideally, we would like to treat the occluders sequentially, maintaining an occlusion cone containing the
effects of all the previously treated occluders. For each occluder, we would retrieve the combined occlusion
cone, blend this cone with the occlusion cone from the occluder, and write the result as the new combined
occlusion cone. Unfortunately, this technique requires the ability to read and write to the same buffer in a
fragment program, or programmable blending capacities. Neither of them are available on current graphics
