52 • CHAPTER 3 Perception
Figure 3.5b shows an example in which the corners and inter-
sections of the fl ashlight’s geons are covered, so the geons can’t
be identifi ed. This is the fl ip side of the principle of componen-
tial recovery—if we can’t see an object’s individual geons, we
can’t recognize the object.
RBC provides an example of bottom-up processing because
its basic unit—the geon—is simple and because perceiving sim-
ple geometric objects like the ones in Figure 3.3 can be related
to patterns of stimulation on the retina. This is similar to how
the cortical neurons in Figure 3.2 can be related to stimuli that
are presented to the retina. But although perceiving objects
begins with stimulation of receptors that leads to the activation
of physiologically or behaviorally determined features, there is
more to perceiving objects than this.Beyond Bottom-Up Processing
If perception were determined solely by bottom-up processing,
then we could understand perception by considering only the
information presented to the receptors. But perception depends
on information in addition to that falling on the receptors,
including knowledge that a person brings to the situation.PERCEPTION DEPENDS ON
ADDITIONAL INFORMATION
Consider the objects in Figure 3.3b. Although the individual
geons that make up these objects may be determined by bottom-
up processing, additional processing is involved when the geons
are combined to create objects. In fact, the same geons can be
combined to create different objects, such as the pail and the
cup. We are able to recognize these different objects based on the
arrangement of their geons, and to give these objects names like
“pail” or “cup,” because of knowledge we bring to the situation.
Processing that begins with a person’s prior knowledge or expec-
tations is called top-down processing. Top-down processing is also involved in our ability
to recognize objects based on just a few geons, as in Figure 3.4, or when large portions of
the object are obscured, as in Figure 3.5. In both of these cases, prior knowledge about
airplanes and fl ashlights probably helps a person perceive these objects.
Another example of how top-down processing is involved in perceiving objects is
illustrated in ● Figure 3.6, which is called “the multiple personalities of a blob” (Oliva
& Torralba, 2007). The blob shown in (a) is perceived as different objects depending on
its orientation and the context within which it is seen. It appears to be an object on a
table in (b), a shoe on a person bending down in (c), and a car and a person crossing the
street in (d). Even though the blob has the same geons in all of the pictures, we perceive
it as different objects because of our knowledge of the kinds of objects that are likely to
be found in different types of scenes.
The idea that perception involves more than bottom-up processing also becomes
apparent when we return to our discussion of physiology. We saw that signals traveling
from the receptors to the brain provide information about an object’s basic features.
However, as these signals travel to the brain, other signals in addition to those generated
by the object’s features become involved as well. Some signals provide information about
other parts of the scene. For example, signals from the tree (green arrows in ● Figure 3.7)
are accompanied by signals from the grass surrounding the tree and from the sky in● (^) FIGURE 3.5 (a) It is possible to identify this object as
a fl ashlight, even though it is partially obscured, because
it is possible to perceive its geons. (b) When the shading is
arranged so the geons can’t be perceived, it is not possible
to recognize the fl ashlight. (Source: From I. Biederman, Computer
Vision, Graphics and Image Processing, pp. 29 and 32. Copyright © 1985 by
Irving Biederman. Academic Press, 1985. All rights reserved. Reproduced
by permission of Elsevier, Ltd.)
(a)
(b)
Copyright 2011 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part.