January, 1993
The computer graphics community is approaching an important threshold, the advent of photorealistic image synthesis available on the desktop. The advances of workstation technology and of basic image synthesis research have combined to allow affordable photorealism: the ability to produce pictures comparable in quality to those captured by a camera.
In issuing its RenderMan 3-D Scene Description Interface proposal, Pixar is embarking on a major effort to make the technology for photorealism accessible for a broad range of scientific visualization and computer-aided-design applications.
The technology for producing interactive display of 3-D graphics has made giant strides in the last decade. Ten years ago, the industry produced 3-D line drawing systems capable of displaying real-time wire-frame renditions of objects. Those systems were put to use across the spectrum of design applications and scientific display problems.
Today, there are graphics workstations capable of real-time shaded-surface display. They are commonly referred to as Gouraud-shaded or Phong-shaded polygon engines. In 3-D CAD applications, the line drawing machines have been supplanted by the newer hardware capable of interactive surface rendering. This is a simple and direct result of the ability of surface representation to convey more information than wire-frame representations.
This technological leap is commendable, but it is just one step. There is no reason to believe that this newer technology is anything more than a temporary plateau, one technological landmark on the journey from wire-frame display to photorealism. The goal in 3-D graphics technology, the ultimate plateau, is the interactive display of photorealistic renditions of designs.
The computer system vendors are providing much higher general computing power for graphics applications. They are looking beyond Gouraud shading. Harnessing the continuing hardware advances to accommodate photorealism is a major challenge to the graphics industry. Forward-looking graphics standards for assimilating these architectures must accommodate photorealism.
The process of creating pictures with a computer generally can be divided in two: modeling and rendering. Modeling serves to describe three-dimensional designs to the computer system. Modeling also includes dynamics, the control of the movement of objects with respect to the simulated camera. Rendering involves the accurate display of the models, subject to their dynamics. Computation of moderate quality renderings is often done in real time, at 30 or 60 frames per second; computation of the highest quality is done off-line, often taking 2 minutes to 2 hours per frame.
Developments in line drawing and shaded surface display technology were preceded by many years of research in the 1960s and 70s. Investigation of methods to heighten the realism of the displayed image was a major research topic in the SIGGRAPH community over the past five to ten years. The industry's most advanced software systems for rendering high-quality images have now largely stabilized. While certain rendering issues, such as radiosity, are the subject of continuing research, far more attention has been paid recently to modeling and dynamics, particularly with regard to natural phenomena and scientific visualization.
Much of computer-aided design of mechanical parts is undertaken by users interested in an accurate physical representation of the part being designed. Photorealism is not the priority. In the creation of engine parts, for example, common concerns are stress and the center of gravity. The early demand for photorealism in CAD comes from users whose design depends more on an accurate and complete visual representation of the part.
The importance in making this technology accessible comes from extending CAD tools into new areas not yet fully addressed with current systems. Architects require pictures involving marble floor, wood paneling, with sunlight streaming in. Automobile stylers require pictures involving chrome, glass, wax, metallic paint, reflections of the great outdoors. Product designers require pictures involving a wide range of shapes, materials, texture. There are other demands outside of traditional CAD markets. Broadcasters require animated sequences of recent and predicted weather patterns and accurate recreations of news events. Scientists require realistic graphics simulations of experimental processes.
Achieving photorealism demands greater sophistication in both modeling and rendering. Most current modeling systems are concerned only with geometry, with the shape of the design. The crucial missing elements are visual attributes, the surface characteristics of the object.
A full model must specify all those elements which contribute to or constrain the final color at every point in a picture: surface color, the location, color and extent of light sources, material translucency and color, pictorial and geometric textures, reflectivity, atmospheric effects. In fact, the computational burden in rendering photorealistic images is in the shading calculations; modeling the shade becomes as important as modeling the shape. Providing usable systems for modeling the surface attributes and complex lighting conditions is a major problem for CAD vendors.
Let us consider the situation as described: graphics workstation performance is just now making photorealism affordable; rendering technology has stabilized; the demand for quality images is known; CAD vendors need a coherent mechanism for modeling visual attributes. The time is right for a 3-D scene description standard for photorealism.
A 3-D scene description interface is the link between a modeling system and a renderer. It is a tool, a language for encapsulating the entire specification of scene information, so that modeling systems have a single fixed point of contact with rendering systems.
If such an interface earns its place as a standard, it improves the efficiency of our industry. Companies building graphics workstations and rendering systems benefit by using the standard as a specification in the design of future systems. With the assurance that rendering systems will grow to meet the standard, CAD vendors benefit as they enrich their tools for specifying visual attributes to address new applications for CAD. Customers benefit because of the availability of richer modeling systems across a range of platforms. The industry grows more efficiently. Agreement on the interface will allow us all to concentrate separately on the important problems of building the needed sophistication into our modeling and rendering subsystems.
The lessons of PostScript® serve well to underscore the potential of such a standard. The existence of PostScript, available with the affordable Macintosh® and LaserWriter® led directly to the phenomenal breakthrough of desktop publishing in the mid-1980s. Various page layout systems (modelers) were developed at that time to drive PostScript, confident that interpreters (renderers) would be available on faster, better, cheaper printing devices.
Pixar envisions nothing less than the equivalent of a LaserWriter for photorealism. The likely outcome will be that photorealistic image synthesis will be a standard offering of graphics machines.
The RenderMan interface is a Pixar proposal designed:
RenderMan is an interface through which 3-D geometry and visual attributes are passed. 3-D geometry is expressed as collections of primitives such as polygons, curved surface patches and NURBs. Constructive solid geometry may be used to assemble the primitives. Visual attributes and shading information is specified with a shading language. Modeling systems use the shading language to express custom materials, textures, light sources, and atmospheric effects. The interface provides a framework for a renderer to achieve a range of capabilities, from antialiasing to ray tracing to motion blur.
The design principles were to keep the interface:
The necessary functionality is offered without frills; many user interface considerations are clearly defined as issues for the modeler feeding the interface. The significant advance of incorporating texture mapping and the full generality of the shading language is designed in an elegant fashion.
The scope of the RenderMan interface in specifying complex and realistic 3-D scenes extends far beyond any existing standards. On the other hand, the interface purposely does not address certain major portions of interactive graphics systems: 2D screen space primitives such as text, markers, and patterns, input, events, picking or interaction, segments or window systems. Such issues are beyond the purview of what a modeler conveys to a renderer.
Graphics standards such as PHIGS and PHIGS+ address the larger problem of an interactive graphics environment. Vendors of 3-D graphics workstations offer graphics libraries with similar scope. In 3-D scene description, these libraries commonly allow for the specification of surface geometry and a fixed lighting model.
The RenderMan interface design takes account of the earlier work in specifying PHIGS. It provides a similar set of geometric primitives, and is directly compatible with the NURBs and trim surfaces added by PHIGS+. The RenderMan interface offers far greater range in the specification of surface characteristics, lights, and environmental conditions. It is our expectation that interactive graphics standards and graphics workstation programming environments can drive the RenderMan interface, and may grow to take advantage of the advanced functionality afforded by RenderMan.
The RenderMan interface was designed in mid-1987 by Pixar as a formal specification of its interface between internal modeling tools and the Pixar software rendering system. Version 2.0 of the RenderMan interface proposal was reviewed and critiqued by 20 companies over a six-month period. Version 3.0 of the interface was issued in May 1988, and Version 3.1 containing the RenderMan Interface Bytestream (RIB), was issued in September, 1989. Public support from the majority of those industry partners that reviewed the proposal indicates the significant level of commitment to the effort. Desktop photorealism will be a reality.
RenderMan® is a registered trademark of Pixar.
PostScript® is a registered trademark of Adobe Systems Incorporated.
Macintosh® and LaserWriter are registered trademarks of Apple
Computer.