P3D Re: Dark Vergence & Cross-viewing

[Larry Berlin <lberlin@xxxxxxxxx>]

>Date: Sat, 3 Jan 1998
>From: Andrew Woods writes:
>.........
>I think I understand...  I experienced something probably akin to what you 
>describe when I watched a 3D movie a while back which had quite a lot of
>vertical parallax. ...............  As the room
>lighting dropped I found it harder and harder to fuse the image - until
>the point at which I could not fuse the image at all when the scene 
>was almost totally black except for the single gas lamp in the middle of 
>the scene.
>
>Prof. Ian Howard of Kent University in Ontario, Canada performed some
>experiments which revealed that with larger 3D images (that filled more
>of your field-of-view) a person's eyes would be more willing to correct
>for alignment errors in the image.  In his experiments the alignment error
>was rotation.
>
>I see this as akin to a "fusion magnet".  If the viewed image is large, the
>fusion magnet will be stronger and therefore be more able to excercise those
>eye muscles to obtain image fusion.

****  I like that term, *fusion magnet*. There is another factor working
here. If the error is rotation, the vertical displacement is different at
different points in the image. If it's just vertically shifted, it exists at
all points, if it has both vertical displacement and rotation, the whole
thing is compounded. Part of what happens is that in the fuller image, there
is more for the eyes, and mind to latch onto. The solid information helps
hold the attention because it is more fusible. Rotation would have some
place where vertical misalignment wouldn't exist providing a *fusion handle*
that allows other aspects to be out of alignment, yet be accepted in the
interpretation.

In examining exactly this effect, I've found in a rotated example, that I
can tilt my head and FEEL certain areas of an image get easier to fuse.
(Since I can distinctly feel this difference, I conclude that perfectly
aligned images are the solution to the common *3D headache* so often
mentioned.) Since roated images have different amounts and directions of
misalignment, I can try several positions of tilting my head and observe
where in the whole image the misalignment most closely matches my head tilt. 
This helps sometimes to determine the center of rotation as well as the
amount of rotation.

It's also easier to ignore vertical misalignment where there is a greater
horizontal parallax, especially behind the stereo window. The extremes of
this effect are to be found in side by side freeviewing. With complete
separation of the convergence points one can adapt to relatively extreme
amounts of misalignment. For example, when I get prints back from
processing, I hand hold the pairs and while arranging them for best effect,
they sometimes get way out of alignment. It's amazing how far I can track
and fuse the images before they break up. The same pictures placed in
essentially the same viewing space, with common points very close together,
as in anaglyphic or LCS, or projected presentation, aren't viewable beyond
very tiny variations.

The hand held viewer is a form of side by side viewing and can tolerate a
greater vertical displacement in the alignment. 

The common 3D image contains small amounts of occasional roation, plain
misalignment, lens length differences, and slight vertical displacements of
the camera between L and R. Most is easy to ignore for experienced viewers.
To those with sensitive eyes it means occasional headaches.

>...........
>Restating your point to how I think I understand it "When viewing a 
>stereoscopic image (e.g. on a PC with shutter glasses), fusing an object 
>which projects out of the screen will require the eyes to converge on a 
>point out of the screen while the eyes must remain focused on the screen.  
>How is this different from cross-viewing?"

*****  It is a form of cross-viewing but with a big difference, in an LCS
system, the images are directed to each eye by timing. The starting point
for the relative convergence points centers on the actual screen depth
making it real easy to establish initial fusion. Once there is someting
obvious to fuse, it's easy and *very natural* to maintain fusion while
crossing the eyes to see parts that project out of the screen. It's
relatively simple convergence. 

In freeview crossed viewing, one must separate the vergence points until
matching visual information is found. Once there, the always present small
eye movements have to be coordinated so as to go the right direction if
fusion drifts. For parallel viewers trying crossed viewing, the automatic
correction goes a long ways in the wrong direction resulting in a feeling of
helplessness and frustration and eye stress. Not until the eyes can correct
in the right direction *automatically* does crossed freeviewing become
comfortable.

>...........
>Also, some people find it hard to view stereoscopic images where the
>images come too far out of the screen.  To cross-view a stereo-pair
>it may require the person's eyes to converge closer than their eyes can 
>comfortably converge.

****  Easily solved by backing up and being farther from the images. Hard to
do if one is watching an IMAX movie... perhaps the best seats are the ones
at the very back, farthest from the screen?

Larry Berlin

Email: lberlin@xxxxxxxxx
http://www.sonic.net/~lberlin/
http://3dzine.simplenet.com/


------------------------------