i
i
i
i
i
i
i
i
10
Stereopsis
One of the main things that sets a VR environment apart from watching TV
is stereopsis, or depth perception. Some 50 years ago, the movie business
dabbled with stereopsis, but it has been only in the last few years that it is
making a comeback with the advent of electronic projection systems. Stere-
opsis adds that wow factor to the viewing experience and is a must for any VR
system.
Artists, designers, scientists and engineers all use depth perception in their
craft. They also use computers: computer-aided design (CAD) software has
given them color displays, real-time 3D graphics and a myriad of human
interface devices, such as the mouse, jo ysticks etc. So the next logical thing
for CAD programs to offer their users is the ability to draw ster eoscopically
as well as in 3D. In order to do this, one has to find some way of feeding
separate video or images to the viewer’s left and right eyes.
Drawing an analogy with stereophonics, where separate speakers or head-
phones play different sounds into your left and right ears, stereopsis must
show different pictures to your left and right eyes. However, stereopsis cannot
be achieved by putting two screens in front of the viewer and expecting her to
focus one eye on one and the other eye on the other. It just doesn’t work. You
can of course mount a device on the viewer’s head (a head-mounted display
or HMD) that shows little pictures to each eye.
Technological miniaturization allows HMDs to be almost as light and
small as a rather chunky pair of sunglasses. F i gure 10.1 illustrates two com-
mercially available models that can be used for stereoscopic viewing or as part
of very personal video/DVD players. However, if you want to look at a com-
235
i
i
i
i
i
i
i
i
236 10. Stereopsis
Figure 10.1. Two examples of typical head-mounted display devices for video, VR
and stereoscopic work. Small video cameras can be build into the eyewear (as seen
on right) so that when used with motion and orientation sensors, the whole package
delivers a sense of being immersed in an interactive virtual environment.
puter monitor or watch a big-screen movie, there is no choice but to find a
different approach. To date, the most successful approach is to combine im-
ages for left and right eyes into a single display that can be separated again by
a special pair of eyeglasses worn by the viewer.
Obtaining stereoscopic images or movies requires a different technology
again. Traditionally, stereoscopic images were obtained using a custom-built
camera or adapters that fit onto the front of conventional single-lens reflex
(SLR) cameras. The headset shown on the right of Figure 10.1 has two
miniature video cameras built into it which generate video streams from the
locations of the left and right eyes. So now we have the option of using two
video cameras and a computer to acquire both still and moving stereoscopic
images and movies.
In this chapter, we will give a short explanation of the theory and termi-
nology of stereopsis before delving a little deeper into stereoscopic technology.
Some example programs for stereoscopic work in a VR context will be pr e-
sented in Chapter 16.
i
i
i
i
i
i
i
i
10.1. Parallax 237
10.1 Parallax
There are many things we see around us that give clues as to how far away
something is or whether one thing is nearer or further away than something
else. Not all of these clues require us to have two eyes. Light and shade, inter-
position, texture gradients and perspective are all examples of monocular depth
cues, all of which we detailed in Section 2.1.2. Another monocular depth cue
is called motion parallax. Motion parallax is the effect we’ve all seen: if you
close one eye and mov e your head side to side, objects closer to you appear
to move faster and further than objects that are behind them. Interestingly,
all the monocular cues with the exception of motion parallax can be used
for depth perception in both stereoscopic and non-stereoscopic display envi-
ronments, and they do a very good job. Just look at any photograph; your
judgment of how far something was from the camera is likely to be quite
reasonable.
Nevertheless, a person’s depth perception is considerably enhanced by
having two eyes. The computer vision techniques which we discuss in Chap-
ter 8 show just how valuable having two independent views can be in ex-
tracting accurate depth information. If it were possible to take snapshots
of what one sees with the left eye and the right eye and overlay them, they
would be different. Parallax quantifies this difference by specifying numeri-
cally the displacement between equivalent points in the images seen from the
two viewpoints, such as the spires on the church in Figure 10.2.
Parallax can be quoted as a relative distance, also referred to as the parallax
separation. This is measured in the plane of projection at the display screen or
monitor, and may be quoted in units of pixels, centimeters etc. The value of
parallax separation is dependent on two factors. The first factor is the distance
between the viewed object and the plane of zero parallax, which is usually the
plane of projection. This effect is shown in Figure 10.3. The second factor
concerns the distance between the viewpoint and the display. The effect of
this is shown in Figure 10.4, where the parallax separation at a monitor screen
at a distance s
1
from the viewpoint will be t
1
, whilst the parallax separation
at a large projection screen a distance s
2
from the viewpoint will be t
2
.We
can avoid having to deal with this second influence on the value of parallax
by using an angular measure, namely the parallax angle which is defined as
. This is also shown in Figure 10.4. For best effect, a parallax ( )ofabout
1.5
◦
is acceptable. Expressing parallax as an angle makes it possible to create
stereoscopic images with parallax separations that are optimized for display
on a computer monitor or in a theater.
i
i
i
i
i
i
i
i
238 10. Stereopsis
Figure 10.2. The images we see in our left and right eyes are slightly different, and
when overlapped they look like we have just had a bad night on the town.
The spatial relationship between viewed object, projection screen and
viewer gives rise to four types of parallax. Figure 10.3 illustrates the four
types as the viewed object moves relative to the projection screen and viewer.
The parallax types are:
• Zero. The object is located at the same distance as the projection screen
and so it appears to hover at the screen.
i
i
i
i
i
i
i
i
10.1. Parallax 239
Figure 10.3. Parallax effects: (a) Parallax t is measured in pixel units or inches or
centimeters; d is the eye separation, on average 2.5 in. (64 mm) in humans. (b) Zero
parallax. (c) Positive parallax. (d) Negative parallax. (e) Divergent parallax (does not
occur in nature).
• Positive. The object appears to hover behind the projection plane; this
effect is rather like looking through a window.
• Negative. Objects appear to float in front of the screen. This is the
most exciting type of parallax.
• Divergent. Normal humans are incapable of divergent vision and there-
fore this form of parallax warrants no further discussion.
Normally the value of parallax separation (t) should be less than the sep-
aration between the two viewpoints (d). This will result in the viewer expe-
Figure 10.4. Expressing parallax as an angle. For example, if a monitor screen is at a
distance of s
1
from the viewpoint, the parallax is t
1
.Wehavethesameβ for a cinema
screen at a distance of s
2
from the viewpoint but the parallax distance is now t
2
.In
the extreme case where the projection plane is moved back into the scene a distance
L, the parallax distance there would be P.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.