Graphics Reference
In-Depth Information
sometimes forget that the famous 'wheel of reincarnation' translates as it rotates,
transporting us to unfamiliar technological territory even if we recognize historical
similarities.”
and Perception
The simple lamp at the top of Figure 1.11 conveys both a shape and a design style
in just a few strokes. Henri Matisse's “Face of a Woman,” shown at the bottom
of Figure 1.11, contains no more than 13 pen strokes but is nonetheless able to
convey an enormous amount to a human viewer; it's far more recognizable as a
face than many of the best contemporary face renderings in graphics. This is partly
because of the
uncanny valley
—an idea from robotics [Mor70] that states that as
robots got increasingly humanlike, a viewer's sense of familiarity would increase
to a point, but then it would drop precipitously until the robot was
very
human-
like, at which point the familiarity would rise rapidly above its previous level. The
uncanny valley is the region in which familiarity is low but human resemblance
is high. In the same way, graphics images of humans that are “almost right” are
often described as “creepy” or “weird.” But ignoring this for a moment, there's
another important difference: Matisse's drawing is simple, whereas an enormous
amount of computational effort is expended in making a realistic face rendering.
This is because artists and designers have reverse-engineered the human visual
system to get the greatest effect for the least amount of “drawing budget.” Look-
ing at their work helps us understand that the goal of all graphics is
communi-
cation,
and that sometimes this is best achieved not with realism but with other
means. Auto-repair manuals, for instance, can be illustrated with photos, but the
top-quality manuals are instead illustrated with drawings (see Figure 1.12) that
emphasize important details and elide other details. Which details are important?
That depends on the intent of the person creating the image and on the human
visual system. We know, for instance, that the human visual system is sensitive
to sharp transitions in brightness and is somewhat more sensitive to vertical and
horizontal lines than to diagonal ones; this partly explains why line drawings are
effective, and why one can afford to leave out diagonal lines preferentially over
verticals and horizontals.
Figure 1.11: The lamp, cour-
tesy of Jack Hughes, has just
five strokes. Matisse's “Face of a
Woman” depicts both shape and
mood in just 13 strokes.
In every engineering problem, there's a budget; graphics is no different. You
are limited in graphics by things like the number of polygons you can send to
the pipeline before you have to draw the next frame to display, the number of
pixels that can be filled, and the amount of computation you can afford to do
in the CPU to decide what polygons you want to draw in the first place. Artists
who are drawing something have a similar budget: the amount of effort spent in
placing marks on a page, the time before the scene being rendered changes (you
can't paint a sunset-in-progress at midnight), etc. They've developed techniques
that allow them to convey a scene on a low budget: For instance, contour drawings
work well, and flat fill-in color adds contrast that helps separate individual objects,
etc. We can learn from the artists' reverse engineering of the human visual system
and use their techniques to render more efficiently. And because most computer-
generated images are intended to be viewed by a human, the human brain is the
ultimate measurement tool for what's satisfactory. There's another budget to con-
sider as well: the viewer's attention. Graphics is also limited by the time and effort
Figure 1.12: A repair manual
shows details where needed, but
omits unnecessary material.
