Introduction

``... in 10 years, all rendering will be volume rendering.''

Jim Kajiya at SIGGRAPH '91 [Elv92]

Only time will show if in 2001 all rendering will be volume rendering. But even today one thing can be said for sure: at least a significant part of all rendering already is volume rendering and will be much more so in 2001. Back in 1991, when Kajiya made his statement, it took at least minutes, if not hours to produce volume rendered images. Today it takes only seconds to produce images of comparable quality and as stated in Moore's Law, computer capacity will increase further in an exponential manner. Of course that does not necessarily mean that in some years all volume rendering will be done in real-time, because additional computing capacity is not only used to shorten rendering times, but also to render more complex and more convincing scenes. In volume rendering, images of radiatively participating media are created. Participating media consist of a large amount of small particles, like water droplets, soot or other suspended solids or individual molecules. Light passing through them may be distracted in many different ways. It may be attenuated by absorption, like light passing through smoke. Participating media may also emit light like fire, or light may be scattered by small particles. In clouds, for example, light is scattered at a myriad of small water droplets. Most rendering toolkits or frameworks have evolved over years and rendering systems have been around for almost decades. Because volume tracing has not been in widespread application for such a long time, only few general ideas exist on how to integrate all the different specialized volume tracing applications into one general purpose rendering system or into existing rendering systems. The main purpose of this paper is to outline a general, flexible framework for volume tracing. In the next section some important concepts of volume tracing and its mathematical foundations will be presented. In the 3rd section, the framework, allowing for a flexible combination of different volume data sources, density functions and transfer functions, is outlined. The 4th section briefly describes an implementation of the framework as an integral part of the Advanced Rendering Toolkit (ART for short) and finally results are presented in the 5th section.