Information Technology Reference
In-Depth Information
∗
6
Case Study: Interpreting Visual Scenes
The previous chapter gave examples of the sort of thinking required to solve certain
puzzles. This type of thinking does have some distinctive features when done by
people. It requires concentration and effort; memory aids (like pencil and paper) are
usually helpful; some people are much better at it than others; and some people see
no fun in the activity at all.
In this chapter, the idea of constraint satisfaction is applied to a form of thinking
that seems much more natural and relaxed:
visual interpretation
. This is a type of
thinking that everyone can do to some extent, typically with a lot less conscious effort
than puzzles require.
The first section briefly considers the concept of vision and its connection to
thinking. The following sections examine a sampling of three visual interpretation
tasks: the interpretation of an image of a two-dimensional terrain as seen from
above (section 2); the interpretation of the edges in an image of three-dimensional
polyhedral
objects
(section
3);
and
recognizing
objects
of
interest
in
an
image
(section 4).
6.1 The thinking part of vision
What is vision? In its simplest terms, it is the process of identifying the physical
objects around us by interpreting the patterns of light that reflect off them. Digitizing
the problem somewhat, one can rephrase it as follows. Imagine that a certain number
of times per second, our eyes are presented with an
image
, that is, a two-dimensional
grid of picture elements (or
pixels
). The job of vision is to respond to these images
with something like “German shepherd puppy” or “Jackson Pollock painting” or
“delivery truck coming straight for me.” The exact details of the image and the desired
responses need not concern us.
