Speaker: Fredo Durand (LCS Graphics Group, MIT)
Visibility problems are central to many computer graphics applications. The most common examples include hidden-part removal for view computation, shadow boundaries, mutual visibility of pairs of points, etc. In this document, we first present a theoretical study of 3D visibility properties in the space of light rays. We group rays that see the same object; this defines the 3D visibility complex. The boundaries of these groups of rays correspond to the visual events of the scene (limits of shadows, disappearance of an object when the viewpoint is moved, etc.). We simplify this structure into a graph in line-space which we call the visibility skeleton. Visual events are the arcs of this graph, and our construction algorithm avoids the intricate treatment of the corresponding 1D sets of lines. We simply compute the extremities (lines with 0 degrees of freedom) of these sets, and we topologically deduce the visual events using a catalogue of adjacencies. Our implementation shows that the skeleton is more general, more efficient and more robust than previous techniques. Applied to lighting simulation, the visibility skeleton permits more accurate and more rapid simulations. We have also developed an occlusion culling preprocess for the display of very complex scenes. We compute the set of potentially visible objects with respect to a volumetric region. In this context, our method is the first which handles the cumulative occlusion due to multiple blockers. Our occlusion tests are performed in planes using extended projections, which makes them simple, efficient and robust. In the second part of the document, we present a vast survey of work related to visibility in various domains.