!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
Kim S. Petersen & Bjarke Jakobsen, IMM, DTU
In Real-time rendering, realistic shadows are important in order to give the viewer a good impression of relations between objects in the scene. To simulate shadows correctly it is essential to have knowledge of visibility between shadow receivers and light sources.
The proposed method provides an approximate visibility function with the ability to work per pixel which is essential in order to produce visually smooth penumbra regions. The visibility function describes how much of an area light source is visible for a given point on a shadow receiver. The method is based on distances between the light source and objects in the scene. These distances are stored in a D-Buffer which can be seen as an augmented shadow map, where euclidian distances are stored instead of depth values. Because of the advanced texturing capabilities provided by modern graphics hardware, the method is implementable on current hardware, giving rise to real-time performance.
Thomas Rued, Digital Arts
Through the combination of 3D hardware and the patented stereoscopic system - ColorCode 3-D, develpers now have an easy way to benefit from stereoscopic rendering in realtime. ColorCode 3-D is the only system in the world that combines full color stereoscopic rendering with the cheap and widely used anaglyh system. With the appearance of 3D rendering hardware - ColorCode encoding can now be done as an integrated part of the traditional renderingpipeline, making it easy for developers to offer good quality stereoscopic features to users of there system.
This sessions is for people already using stereo rendering in there work, people considering what solution is the best for an upcoming stereo project, or for everybody else interested in stereoscopic rendering. It shows how a modern 3D engine can benefit from this extra dimension in an easy and usable way.
Andreas Bærentzen, IMM, DTU.
Point Rendering is currently becoming more popular. This is probably due to several things. For instance, model sizes grow, and for very large triangle mesh models, the size of a triangle projected onto the screen can easily be smaller than a pixel.
In this talk, we propose the use of points as a technique for impostor rendering. An impostor is a simplified model of complex geometry. We first generate a set of points from a triangle model. This is done using a technique called depth peeling. Depth peeling generates a Layered Depth Image (LDI) representation of an object. From the LDIs we extract a set of points. We show how a random ordering of these points can be used as a continuous LOD representation. In fact, the random permutation of points seems to be the simplest conceivable continuous LOD representation.