Graphics Reference
In-Depth Information
considered as computer graphics, but disappointing when considered as art: They
managed to capture only the surface veneer of the art. Nonetheless, there was
considerable value in this work, not only as a foundation for later work where
intent and abstraction were incorporated, but in immediate applications as well.
For instance, rendering with a rough pencil-sketch appearance conveys implicitly
the idea that the rendering (or the thing being rendered) is incomplete, or that
details are not important. A user-interface mockup drawn in a pencil-sketch style
for initial testing can help get users to say, “I really need a brightness knob,” rather
than “I don't like the glossy highlights on the knobs,” for instance.
Before discussing style and abstraction further, we'll examine some early exam-
ples of expressive rendering (see Figure 34.4). These use various kinds of input,
from imagery to purely geometric models to human-annotated models.
The first example comes from the work of Saito and Takahashi [ST90], who
recognized that in the course of rendering an image, one could also record at each
pixel a depth value for the object visible at the pixel, or the texture coordinates on
the object at that point, or any other property. Using image-processing methods to
detect discontinuities in depth allowed them to extract contours of shapes; simi-
larly, searching for derivative discontinuities allowed them to detect edges (like the
edges of a cube). By rendering these contour and edge pixels in highlight colors
over the original, they created renderings that they characterized as “comprehen-
sible,” indicating their belief that the additional lines helped the visual system to
better understand the thing being seen. The second example—a pen-and-ink ren-
dering made from a more complex model—shows the application of
indication,
a technique in which something recurrent (like the pattern of shingles on a roof,
or bricks in a wall) is suggested to the eye by just drawing a small portion of it.
The third shows both the visible and the hidden contours of a polyhedral model;
these have been extracted at real-time rates and assembled into long arcs, and then
these arcs have been rendered with a “style” giving a richer appearance than a
simple pen stroke. The fourth shows an example of stylistic imitation: The render-
ing starts from geometric models that have been enhanced with finely randomly
sampled points; attached to each point is a brushstroke (one of several scanned
images of actual oil-paint strokes) and a color (determined by a reference image,
a lighted and shaded rendering of the original scene). Rendering consists of draw-
ing (i.e., compositing into the final image) the strokes in a back-to-front order.
There are many details remaining (the orientation of strokes, the creation of refer-
ence images, etc.), but the essential result is to give the appearance of a painting.
If the strokes are similar to Monet's, and the reference image's coloring is similar
to Monet's, and the chosen scene is similar to something that Monet might have
painted, the final result will resemble a Monet painting.
Because expressive rendering is comparatively new, the overarching principles
and structures for the area have not yet become apparent. The remainder of this
chapter therefore consists of some general material that applies to enough different
techniques that it deserves discussion, and a tour of some specific techniques that
we think illustrate various important points, followed by some brief conjectures
about future directions and related work.
