SIGGRAPH is the most significant conference in computer graphics. It brings together game developers, film industry professionals, and scientists for a week in August. They present today's hottest products in each field and the new research techniques that will drive tomorrow's development. This article reviews the research papers presented this year that are the most relevant to game developers. SIGGRAPH also contains an exhibition, courses, technical sketches, animation festival, and special events.

However, one of the more subtle ways that the original Star Wars film changed science fiction forever was. dirt. Star Wars was the first film where spaceships and technology were covered in oil and grid instead of appearing as fresh, gleaming objects. Rust, stains, and dirt make objects appear part of their environment and suggest a history that began long before the observer arrived. Lucas points out that the more fantastic the story and setting, the more important it is to make it familiar and realistic so that the audience can relate.

This applies to games as well. Good 3D game artists create meshes and textures that are rusty and dinged-up like the spaceships in Star Wars. This year, Chen et al. introduced a new algorithm for helping with this process called Visual Simulation of Weathering by Gamma-ton Tracing. The "gamma-tons" are imaginary old-age particles that fall from the sky and bounce around a scene. Wherever they contact objects they induce aging artifacts like rust. Because the path of each gamma-ton is traced they are good for simulating effects like stains and moss-growth that flow outward from a source. The ideal use for gamma-tons in games is as part of a level compiler. After building a clean scene, the level compiler can crunch on not only light-maps and AI planning but on weathering the environment for realism.

Game modeling is a constant tension between adding detail like dings and cracks, and removing polygon detail to make scenes animate and render faster. Lee et al. presented a paper that improves detail reduction. It is based on the idea of saliency maps, which describe the relative impact of different parts of an object to a human observer's perception. Previous salience algorithms have been directed at images. There, features like line intersections, text, and human faces are marked as important (salient). Lee et al.'s Mesh Saliency advances the state of the art by describing the first algorithm for computing saliency on a 3D mesh. This is exciting news for artists. The result opens the possibility of automatic LOD algorithms that reduce polygon count while maintaining the perceived shape of an object. In the paper, the authors give examples of automated mesh simplifications that appear significantly better than results produced with conventional algorithms. Programmers should be able to directly implement the paper's methods for in-house use (particularly for normal-map generators) and hopefully salience-based simplification will soon appear in modeling tools.