ABSTRACT
Even though shadow algorithms have been around for almost as long as computer graphics itself, robust and e›cient hard-shadow generation is still not a solved problem. While geometry-based algorithms produce pixel-perfect results, they sušer from robustness problemswith dišerent viewer-light congurations and are oŸen slow due to the enormous overdraw involved in rasterizing shadow volumes. Shadow-map algorithms, on the other hand, are very fast as their complexity
is similar to standard scene rendering, but they sušer from aliasing artifacts since the sampling of the shadow map and the sampling of the image pixels projected into the shadow map usually do not match up. In this chapter, we will analyze aliasing in more detail. We rst show the dišerent components of aliasing in a signal-reconstruction framework (Section 3.1) and then go intomore detail on the principal aliasing components, initial sampling error (Section 3.2) and resampling error (Section 3.3).
