Graphics Reference
In-Depth Information
eventually provided a general ambient light that illuminated things even when
they weren't directly lit, and that the colors at the interior points of any triangle
could be inferred from the colors computed at the triangle's vertices.
Gradually, richer and richer models—of shape, of light, and of reflectance—
were added, but even today the dominant model for describing the light in a scene
includes the term
“ambient,”
meaning a certain amount of light that's “all over the
place in the scene” without any clear origin, ensuring that any object that's visible
in the scene is at least somewhat illuminated. This ad hoc term was added to
address aspects of light transport, such as interobject reflections, that could not be
directly computed with 1960s computers; but it remains in use today. While many
topics follow the historical development of light transport, we'll choose a different
approach and discuss the ideal (the physical simulation of light transport), how
current algorithms approximate that ideal, how some earlier approaches did so as
well, and how the vestiges of those approximations remain in common practice.
The exception to this is that we'll introduce, in Chapter 6, a reflectance model that
represents the scattering of light from a surface as a sum of three terms:
“diffuse,”
corresponding to light that's reflected equally in all directions;
“specular,”
4
used
to model more directional reflection, ranging from things like rough plastic all the
way to the nearly perfect reflection of mirrors; and ambient. We will refine this
model somewhat in Chapter 14, and then examine it in detail in Chapter 27. The
advantage of the early look is that it allows you to experiment with modeling and
rendering scenes early, even before you've learned how light is actually reflected.
Graphics displays have improved enormously over the years, with a shift from
vector devices to
raster
devices—ones that display an array of small dots, for
example, like CRTs or LCD displays—in the 1970s to 1980s, and with steadily
but slowly increasing
resolution
(the smallness of the individual dots), size (the
physical dimensions of the displays), and
dynamic range
(the ratio of the bright-
est to the dimmest possible pixel values) over the past 25 years. The performance
of graphics processors has also progressed in accordance with Moore's Law (the
rate of exponential improvement has been greater for graphics processors than for
CPUs). Graphics processor architecture is also increasingly parallel; how far this
can go is a matter of some speculation.
In both processors and displays, there have also been important leaps along
with steady progress: The switch from vector devices to raster displays, and their
rapid infiltration of the minicomputer and workstation market, was one of these.
Another was the introduction of commodity graphics cards (and their associated
software), which made it possible to write programs that ran on a wide variety
of machines. At about the same time as raster displays became widely adopted
another major change took place: the adoption of Xerox PARC's WIMP GUIs.
This is when graphics moved from being a laboratory research instrument to being
an unspoken component of everyday interaction with the computer.
One last leap is worth noting: the introduction of the
programmable
graphics
card. Instead of sending polygons or images to a graphics card, an application
could now send certain small
programs
describing how subsequent polygons and
images were to be processed on their way to the display. These so-called “shaders”
4. The word “specular” has multiple meanings in graphics, from “mirrorlike” to “any-
where from sort of glossy to a perfect mirror.” Aside from its use in Chapter 6 we'll
use “specular” as a synonym for mirror reflection, and “glossy” for things that are shiny
but not exactly mirrorlike.
