Three-dimensional computer graphics
Three-dimensional computer graphics, sometimes called 3D computer-generated imagery, means graphics that use a three-dimensional representation of geometric data. The coordinates are often Cartesian. The computer stores them so it can calculate, and so it can render a digital image. The image is usually flat. Sometimes it is a stereoscopic pair, or a picture aimed at a display that really does have depth. The result may be stored and watched later, as an animation, or shown as it is computed.
The name suggests a solid picture. What you usually get is a flat one. Unlike a 3D film, where the presentation itself is meant to produce visual depth, a rendered frame on an ordinary monitor has no depth you can measure with a ruler. It has the cues of depth. Increasingly the same data is also sent to a display that does have depth, including the displays in virtual-reality systems. The data did not change category. The glass did.
3D graphics stand apart from 2D computer graphics, which typically use different methods and different formats both to create the picture and to render it. They also rely on a lot of the same algorithms: vector methods in the wire-frame model, raster methods in the final image. In real software the line blurs. A 2D program may use 3D techniques to get a lighting effect. A 3D program may use 2D techniques for a pass that has nothing to do with the camera. The labels on the box are coarser than the code.
A model is not yet a picture
The objects are called 3D models. Unlike the rendered image, a model's data lives in a graphics file. A 3D model is a mathematical representation of a three-dimensional object. It is not technically a graphic until it is displayed. That sounds like a quibble. It saves a lot of confusion. You can run a simulation on a model and never draw it. You can draw it badly and the model is still the model. With 3D printing, the model is rendered into a physical object, with limits on how closely the physical thing can match the virtual one. A printer has a nozzle, a layer height, and materials that sag. The file does not know about any of that until a second program tells it.
The production of a picture falls into three phases.
- Modelling: forming a computer model of the object's shape.
- Layout and animation: placing and moving objects, models, lights, and the rest, inside a scene.
- Rendering: the calculations that, from the lights, the surface types, and the other qualities, produce an image.
Building a model takes the first two. Light and materials takes the third. A scene file is what sits between them: geometry, a viewpoint, textures, lights. Rendering reads that file and writes pixels.
Wire frames, then shading
Sketchpad, demonstrated in 1963, was a program for drawing in two dimensions. The same year, Timothy Johnson's Sketchpad III extended it to interactive wire-frame models in three dimensions, with a perspective view beside the top, the front, and the side. Hidden edges remained visible. You could tell the object was solid only by reading the lines carefully. Shading research through the middle of the 1970s is what made polygonal models look continuous instead of faceted. Arthur Appel, in 1968, set out point-by-point shading of solid objects, together with the problem of deciding which regions lie in shadow. Henri Gouraud's 1971 method computed illumination at the vertices of a polygon and interpolated it across the face, which reduced the visible seams. Bui Tuong Phong's 1975 model interpolated surface normals and evaluated a specular term at each sample, which put a tighter highlight on a polygonal approximation of a curve.
One of the first times a cinema audience saw this kind of animation was Futureworld in 1976, which included a face and a hand first made for the 1972 short A Computer Animated Hand, by Edwin Catmull and Fred Parke, then students at Utah. Home computers got 3D software of a sort late in the same decade. 3D Art Graphics, written by Kazumasa Mitazawa, came out in June 1978 for the Apple II.
Turner Whitted's 1980 illumination model continued rays past the first surface, so reflection, refraction, and shadows from point lights were computed together. He was extending rays cast from the viewer, a method he credited to Appel, rather than rays followed outward from the lamps. In 1984 Robert L. Cook, Thomas Porter, and Loren Carpenter published distributed ray tracing, replacing those exact secondary rays with stochastic samples. Motion blur, depth of field, penumbras, translucency, and glossy reflection are among the effects those samples can produce. James Kajiya's rendering equation, at SIGGRAPH in 1986, writes light transport as one integral equation. Whitted's model and distributed ray tracing are approximations of it. Path tracing, the Monte Carlo solution he described, covers a wider range of optical effects than a fixed tree of reflection and refraction rays.
For feature animation, Cook, Carpenter, and Catmull described the Reyes architecture in 1987. Each surface is diced into micropolygons, shaded in that surface's own coordinates, and tested for visibility by stochastic sampling into a z-buffer. On Silicon Graphics and Sun workstations, artists at Pixar built the models for Toy Story (1995), and RenderMan produced the frames. John Lasseter directed it. It was the first theatrical motion picture whose imagery was entirely computer-generated.
On 31 August 1999 Nvidia unveiled the GeForce 256, one chip integrating transform, lighting, and rendering, which the company called the world's first GPU. The earlier history of graphics chips is a different "first," and it is told with the workstations. A head-mounted display for virtual reality was patented in 1960. The Oculus Rift project began with a Kickstarter campaign in 2012. The gap between those dates is the gap between a patent and a product a person could order.
None of this requires the picture to look like a photograph. A model can be rendered in an art style on purpose. And a picture can look solid without a model at all, which is a different subject: pictures that only look deep.