Two ways to put a picture on a screen
Computer graphics is the use of a computer to create and display images. The field has a few recurring problems: how to represent a picture inside the machine, how to calculate what it should look like, and how to show it on a display. Nearly every stage in its history is an answer to one of these, shaped by what the available hardware could afford.
There are two basic ways to show a picture, and the history is partly the story of the shift from one to the other. In a vector display, the screen’s beam draws lines directly, following coordinates from the computer. In a raster display, the picture is a grid of small dots, called pixels, which the hardware scans through row by row. The early interactive systems of the 1950s and 1960s, described in the article on interactive graphics, were mostly vector based.
Vector displays
Vector screens were well suited to early computers. A picture could be stored as a short list of line endpoints, which took little memory, and the lines were crisp. Their weaknesses were that complicated drawings flickered when the computer could not redraw them fast enough, and that they were poor at filled areas and shades of gray. For a long time, vector systems were the standard for engineering, and some early arcade games used them for their sharp lines. The technology behind these screens is described in the article on CRT displays.
Wireframes and the problem of 3D
Showing a three-dimensional object on a flat screen requires converting its coordinates into a two-dimensional view, a calculation known as projection. In the 1960s researchers wrote programs to rotate and display wireframe models, which show only the edges of an object. A difficulty emerged immediately: if all the edges are drawn, the viewer sees through the object, and it is hard to tell front from back. Solving the hidden-surface problem became one of the central research topics of the late 1960s and 1970s, and several different algorithms were proposed by various groups.
The University of Utah, with support from US government research funding, became a center for this work in the late 1960s and the 1970s. Its students and staff contributed ideas that spread throughout the industry. A teapot model created there in the mid-1970s remains a well-known test object.
Raster images and frame buffers
The raster approach needs memory to hold the colors of every pixel, a block of memory known as a frame buffer. In the 1960s and early 1970s such memory was very expensive. As the cost of memory fell, systems with frame buffers appeared in research labs in the 1970s, including at Xerox PARC, where raster displays were part of the development of the first personal workstations. They could show filled areas, shades and, later, color, and they were soon able to display text in different fonts and windows, which suited the graphical user interface.
As memory prices continued to fall, raster displays became the norm for televisions, computers and games. Early home computers and consoles typically had very limited resolution and few colors, and artists learned to work within those limits, a craft described in the article on pixel art.
Shading, texture and light
Making a raster picture of a 3D object look solid required more than hidden-surface removal. During the early 1970s, researchers proposed methods of shading, in which the brightness of a surface changes smoothly across a model, rather than appearing as flat facets. Two such methods developed at Utah, by Henri Gouraud and by Bui Tuong Phong, are still taught as the standard introductions. Around the mid-1970s, Edwin Catmull’s work introduced texture mapping, which wraps a picture over a model’s surface, and the depth buffer, which stores the distance of each pixel so that nearer objects cover farther ones.
Around 1980, Turner Whitted published a paper describing ray tracing in a form that included reflections and transparent materials. In ray tracing, the program follows imaginary rays of light back from the viewer through the scene. It produced realistic images but needed large amounts of calculation. Studios used rendering of these kinds for film, as described in the account of computer-generated imagery, but each frame could take a long time.
Standards, workstations and the graphics card
In the 1980s, graphics workstations became widely used in engineering, science and film. Their manufacturers developed programming interfaces, and this led to shared standards. OpenGL, introduced in the early 1990s, gave programmers a common way to draw 3D images on different hardware.
At the same time, home computers and consoles gained dedicated graphics circuits. By the second half of the 1990s, PC users could buy add-in cards that took over the calculations for drawing triangles, and by the end of the decade the term graphics processing unit was being used for chips doing a wide range of graphics work. Over time these processors became programmable, and they were applied to tasks other than graphics, including scientific computing. The expansion in power also supported game engines, which package rendering and other functions for reuse.
What the history shows
The history of computer graphics follows a clear pattern: ideas are often published long before the hardware exists to run them in real time. Ray tracing, for instance, was known in the 1970s but did not become practical for interactive use until much later. Hardware cost, especially the cost of memory, repeatedly decided which approach prevailed.
Dates and credit in this field are often given with more certainty than the record supports, since similar ideas appeared in parallel at several institutions. The sequence above is a general outline, intended to explain how the technical concepts built on one another.