i
i
i
i
i
i
i
i
22.5. Picture Perception 593
we interpret what we notice (Pashler, 1998). Figure 22.43 provides an example of
how attentional processes affect vision, even for very simple images. In the left
two panels, the one pattern differing in shape or color from the rest immediately
“pops out” and is easily noticed. In the panel on the right, the one pattern differ-
ing in both shape and color is harder to find. The reason for this is that the visual
system can do a parallel search for items distinguished by individual properties,
but requires more cognitive, sequential search when looking for items that are in-
dicated by the simultaneous presence of two distinguishing features. Graphically
based human-computer interfaces should be (but often are not!) designed with an
understanding of how to take advantage of visual attention processes in people so
as to communicate important information quickly and effectively.
22.5 Picture Perception
So far, this chapter has dealt with the visual perception that occurs when the world
is directly imaged by the human eye. When we view the results of computer
graphics, of course, we are looking at rendered images and not the real world.
This has important perceptual implications. In principle, it should be possible to
generate computer graphics that appears indistinguishable from the real world, at
least for monocular viewing without either object or observer motion. Imagine
looking out at the world through a glass window. Now, consider coloring each
point on the window to exactly match the color of the world originally seen at
that point.
10
The light reaching the eye is unchanged by this operation, meaning
that perception should be the same whether the painted glass is viewed or the
real world is viewed through the window. The goal of computer graphics can be
thought of as producing the colored window without actually having the equiva-
lent real-world view available.
The problem for computer graphics and other visual arts is that we can’t in
practice match a view of the real world by coloring a flat surface. The brightness
and dynamic range of light in the real world is impossible to recreate using any
current display technology. Resolution of rendered images is also often less that
the finest detail perceivable by human vision. Lightness and color constancy are
much less apparent in pictures than in the real world, likely because the visual
system attempts to compensate for variability in the brightness and color of the
illumination based on the ambient illumination in the viewing environment rather
than the illumination associated with the rendered image. This is why the real-
10
This idea was fi rst described by the painter Leon Battista Alberti in 1435 and is now known as
Alberti’s W indow. It is closely related to the camera obscura.
i
i
i
i
i
i
i
i
594 22. Visual Perception
istic appearance of color in photographs depends on film color balanced for the
nature of the light source present when the photograph was taken and why real-
istic color in video requires a white-balancing step. While much is known about
how limitations in resolution, brightness, and dynamic range affect the detectabil-
ity of simple patterns, almost nothing is known about how these display properties
affect spatial vision or object identification.
We have a better understanding of other aspects of this problem, which psy-
chologists refer to as the perception of pictorial space (S. Rogers, 1995). One
difference between viewing images and viewing the real world is that accommo-
dation, binocular stereo, motion parallax, and perhaps other depth cues may indi-
cate that the surface under view is much different that the distances in the world
that it is intended to represent. The depths that are seen in such a situation tend
to be somewhere between the depths indicated by the pictorial cues in the image
and the distance to the image itself. When looking at a photograph or computer
display, this often results in a sense of scale smaller than intended. On the other
hand, seeing a movie in a big-screen theater produces a more compelling sense of
spaciousness than does seeing the same movie on television, even if the distance
to the TV is such that the visual angles are the same, since the movie screen is
farther away.
Computer graphics rendered using perspective projection has a viewpoint,
specified as a position and direction in model space, and a view frustum, which
specifies the horizontal and vertical field of view and several other aspects of the
viewing transform. If the rendered image is not viewed from the correct location,
the visual angles to the borders of the image will not match the frustum used in
creating the image. All visual angles within the image will be distorted as well,
causing a distortion in all of the pictorial depth and orientation cues based on
linear perspective. This effect occurs frequently in practice, when a viewer is po-
sitioned either too close or too far away from a photograph or display surface. If
the viewer is too close, the perspective cues for depth will be compressed, and the
cues for surface slant will indicate that the surface is closer to perpendicular to the
line of sight than is actually the case. The situation is reversed if the viewer is too
far from the photograph or screen. The situation is even more complicated if the
line of sight does not go through the center of the viewing area, as is commonly
the case in a wide variety of viewing situations.
The human visual system is able to partially compensate for perspective dis-
tortions arising from viewing an image at the wrong location, which is why we
are able to sit in different seats at a movie theater and experience a similar sense
of the depicted space. When controlling viewing position is particularly impor-
tant, viewing tubes can be used. These are appropriately sized tubes, mounted
i
i
i
i
i
i
i
i
22.5. Picture Perception 595
in a fixed position relative to the display, and through which the viewer sees the
display. The viewing tube constrains the observation point to the (hopefully) cor-
rect position. Viewing tubes are also quite effective at reducing the conflict in
depth information between the pictorial cues in the image and the actual display
surface. They eliminate both stereo and motion parallax, which if present would
correspond to the display surface, not the rendered view. If they are small enough
in diameter, they also reduce other cues to the location of the display surface by
hiding the picture frame or edge of the display device. Exotic visually immersive
display devices such as head-mounted displays (HMDs) go further in attempting
to hide visual cues to the position of the display surface while adding binocu-
lar stereo and motion parallax consistent with the geometry of the world being
rendered.
i
i
i
i
i
i
i
i
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.