(c)  Copyright 1989 Commodore-Amiga, Inc.   All rights reserved.
The information contained herein is subject to change without notice, and 
is provided "as is" without warranty of any kind, either expressed or implied.  
The entire risk as to the use of this information is assumed by the user.



                       STEREOSCOPY AND THE AMIGA

                          by Tracy McSherry


Visual information comes to the brain from over seven million rods and 
cones in the eyes that receive light through an adjustable lens and 
aperture system.  All this information is processed by the brain in 
real-time to produce an integrated image.  Visual depth perception
is part of that processing.  Depth perception is achieved through 
visual cues intrinsic to a binocular visual system.  


DEPTH PERCEPTION

These depth cues include Convergence, Binocular Disparity, Accomodation, 
Motion Parallax and Pictorial Clues.  

o  Convergence is a depth cue from the interocular angle formed by the
   eyes looking at an object.  Through trial and error over many years 
   you have learned to estimate the distance of close objects by 
   triangulation and feedback from the muscle tension and eye position 
   of crossed eyes.

o  Binocular Disparity is the variance in images caused by the different 
   path of light entering each eye.  Slightly different angles, sizes, 
   positions, brightness and color values all result from the displaced 
   geometry of each eye.  The two images are compared and fused and the 
   difference is interpreted to give a depth perception.

o  Accommodation refers to the adjustment in focus by the lens
   muscles straining on the lens.  Here, feedback that results in a
   sharp image can also give depth information.  This doesn't get
   used much in computer displays since the image plane is constant - 
   the monitor doesn't move. 

o  Motion Parallax is a cue from the angular relative motion that is
   affected by distance.  The best example of this is driving down
   the street and seeing the stripes race by while the far away
   objects appear to move slower.  Motion Parallax provides a very good 
   3D effect on a computer display and does not require 3D glasses.

o  Pictorial Clues include matting or overlapping, shading, horizons, 
   perspective, scaling, texture, blurring, contours and shadows.  
   Pictorial Clues also work well in computer applications and do not
   require special glasses.

You can demonstrate the effect of some of these depth cues on yourself.
For instance, hold your finger straight up in front of your face.  Now 
close your left eye keeping your right eye open.  Switch eyes and 
your finger should appear to jump.   The farther away the object, the 
smaller the jump.  This is called parallax - the separation of the eyes 
gives each a slightly different image and perspective.  

By trigonometry we know that given one side and two angles, any triangle 
can be defined.  The distance between a person's eyes remains constant,
so the brain learns to use the relative angles of the eyes as a clue to
how far away an object is - this is the convergence depth cue.

For instance, when you look at an object which is very far away, there 
is no pronounced difference between the two images from each eye.  The 
interocular angle is close to zero and the eyes look out almost parallel 
to each other.  The brain uses all this information in gauging depth.

Here is another example you can try.  Hold one finger up at arms length 
and the other halfway in between.  If you look at one finger, you see 
two images of the other.  Since the eyes can only focus and cross at one 
distance at a time, anything out of focus must be at another depth.  The 
brain uses this as another depth cue.

Another example you can try is called the "floating finger".  Put your 
hands one foot in front of your face with your index fingers pointing at 
each other.  Now look out at something past the tip of your fingers, at
least three feet away.  If you focus on the distant object you should see 
a third finger "floating" between the other two.  Of course you can
concentrate and ignore one image over the other or shift focus and resolve 
one image.  But this shows how easily the brain integrates dual images, 
sometimes producing a new object that isn't really there.  By taking 
advantage of the how the brain processes visual data, we can produce 
realistic 3D effects on the Amiga.


3D EFFECTS ON THE AMIGA

Three dimensional imaging is implemented on the Amiga by creating two 
separate views for each eye.  The two images are displayed alternately 
with the Amiga's interlace mode.  The display is viewed through a pair 
of stereoscopic glasses such as the LASER GAMESMANSHIP STEREOSIGHT GLASSES, 
the HAITEX X-SPECS 3D visor, or a suitably modified pair of SEGA 3D glasses.  

Using a field sequential approach, the liquid crystal shutters in the glasses 
separate the interlaced display into a left view and a right view.  The 
liquid crystal shutters are synchronized with the interlace screen so that 
each eye sees a separate field - a noninterlaced medium resolution picture.  

The brain integrates the two views, merging together the objects identified 
as the same object by crossing the eyes until the overlap makes a single 
object.  Since the eyes are crossed either more or less than they would be
normally when looking at the screen, the stereo image appears either closer
or farther away.  

The stereo objects that merge also seem to be more in focus.  Other objects 
remain doubled unless they have the same separation or effective parallax.  
The brain assumes these other objects are out of focus and ignores them the 
way it is used to doing.  Some people don't expect this to happen on the 
computer screen when there are two objects at different depths.  

Note however that the eyes focus on the screen surface as a single image 
plane.  Since the computer screen does not move, the eyes will not change 
focus as you look from one object to another.  Hence, Accomodation cues 
cannot be simulated.  With stereo imaging we can only play with the 
convergence.  


SIMULATING CONVERGENCE

Independent control of what each eye sees makes it possible to simulate 
convergence.  To see how this works, consider a few examples.  Think in 
terms of a triangle with the base of the triangle the line between your 
eyes. (Hopefully we can take that distance as a constant!)  If you don't 
move your head, the distance to the monitor is constant.  Assume that the 
bridge of your nose is directly in front of the center of the monitor
as in Figure 1.  In this case the image appears at the same depth as the 
screen,  because both eyes see an identical image and the visual cues
correlate and reinforce the depth perception.

 
----------------------INSERT FUGURE 1 HERE-------------------------


By offsetting the two images so the eyes cross as in Figure 2, the 
convergence indicates that the object is at the point where the eyes cross.  
In this case the virtual image appears closer than the screen.  This can be 
reinforced by giving a different view or perspective of each stereo image.  
This is a good effect but crossing the eyes too severely, too often will 
cause the eye muscles to strain and should be avoided.

 
----------------------INSERT FUGURE 2 HERE-------------------------


More common and easier on the eye is the opposite separation as in Figure 3 
where the eyes cross past the plane of the screen.  In this case the virtual 
image appears farther away than the screen.  
 

----------------------INSERT FUGURE 3 HERE-------------------------


At the extreme, if the screen images are almost six centimeters apart, 
the eyes are looking straight ahead as in Figure 4.  In this case the 
virtual image appears at infinity.

 
----------------------INSERT FUGURE 4 HERE-------------------------




ON MAKING 3D IMAGERY

The nice thing about doing stereoscopic images on the AMIGA is the number 
of 3D rendering packages.  All of them can be used to generate stereoscopic 
images, by moving the camera or the observers position.  Unfortunately none 
of them currently fully support stereo imaging and so they allow you to make 
embarrassing mistakes. 

To get the best results, the camera movement should be small and understated 
to avoid inducing eye strain or confusing the viewer.  Excessive parallax 
will prevent comfortable viewing and is usually unnecessary.  

Use stereo effects with discretion.  If you did a stereoscopic drawing 
right now, would you put everything in the drawing in 3D?   You only want 
to make the focal point of the picture finely detailed - this could include
stereo effects but don't apply 3D to everything just because you've paid 
fifty dollars for a pair of glasses.

Proper use of stereo images can make a complex screen display appear 
simple by allowing the eyes to look at it a layer at a time. But 
remember, there is no magical way to increase the amount of information 
the interlaced display puts out, so keep in mind that an increase in depth 
perception costs in detail or resolution.  Conversely we can provide more 
information at different depth levels if detail is not important.

For CAD and circuit layout, depth information is critical and can make life 
much easier.  For a nice picture of an orange, color and resolution are more 
important.  To create realistic action and interactive entertainment, a 
compromise is in order.

When doing three dimensional or stereo displays you have to take care to 
avoid inducing eye strain or confusing the viewer.  Try crossing your eyes 
and then looking at something.  You'll notice a fairly long period of 
unfocused viewing.  You don't want to quickly change the separation of 
your stereo images or you might induce the same effect.

Avoid conflicting information.  Big cars and small buildings will do more 
to give wrong information than you can compensate for by stereo images.  
It will always seem that the cars are closer than you want them to be.  
Putting an object in front of another object means it is closer, no 
matter what other clues you add.

A temptation exists to try to use stereo pairs as the only clue to depth.  
You need to remember that although you can get depth information from 
many different types of cues, the best perception of depth is achieved
when several are used in conjunction.  Size, shading, perspective and 
peripheral clues are just as important.  Careful use of each will
maximize the effect and really enhance your applications.

With careful manipulation of objects and experimentation, you can create 
stunning stereoscopic images and given, the power of the AMIGA, even 
stereoscopic animations and games.  

