Extensible Image Synthesis

Abstract

Currently the two major obstacles for the limited growth of 3D application development to become a mainstream technology for everyday use are a) the computational and rendering requirements of 3D and b) the lack of a programming model that is appropriate for widespread use by developers who are not experts in the field of 3D graphics. But as the hardware gets faster, software will become the critical factor in the further growth of 3D application development. In this paper we present a software architecture for a 3D rendering package which can operate in one of two modes: it either ray-traces the scene, creating photorealistic images, or it transforms the scene into a set of graphics primitives typically supported by a standard graphics package. The renderer is object-based rather than drawing based and consists of an extensible set of objects that perform a variety of operations. The 3D objects as well as the imaging objects (like Image, Screen, Light) are the building blocks that lend themselves to programmer customization through techniques such as subclassing. State-of-the-art functionality and advanced algorithms can be incorporated into this renderer with a minimum amount of programming (i.e. Analysis/understanding of existing code and creation of new code).

A thorough test of this approach has been carried out by using the renderer as the platform for teaching and for lab assignments in several undergraduate and graduate courses at two different universities. Experiences with this (inhomogeneous) user population prove that the system meets its design goal of being highly customizable and extendable.