[opendtv] The Plain Geometry of Stereoscopic 3D
- From: Mark Schubin <tvmark@xxxxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sun, 07 Mar 2010 10:22:38 -0500
I've changed the title of the thread to reflect your question.
There are three issues related to screen size and distance.
One is human variation in interpupillary distance (IPD). When we look
at objects at an infinite distance, our eyes point straight ahead, and
even when looking only about 30 feet away we're pointing almost straight
ahead. So, when showing something of infinite distance on screen, the
left- and right-eye views should be spaced apart as much as our eyes.
Unfortunately, our eye-separation distances vary.
If you look up "Interpupillary Distance" on Wikipedia, you'll find a
1988 survey showing the mean adult male IPD as 64.7 mm. That's the 65
mm to which Jeroen & I referred or the roughly two-and-a-half inches
bandied about. But the same survey's mean adult female IPD was 62.3 mm,
and the adult range was from 52 to 74 mm
If, however, you look up "Pupillary Distance" on Wikipedia, you'll find
somewhat different numbers, between 48 and 73 mm for adults and between
41 and 55 mm for children in the U.S., with female distances somewhat
smaller than male and, according to that source, with Europeans slightly
smaller than Americans. This 2004 paper has adult males as wide as 78
mm and infants as low as 32 (with a possible racial variation):
http://www.cl.cam.ac.uk/~nad10/pubs/EI5291A-05.pdf
That's why binoculars and binocular microscopes have adjustable IPDs and
why those dealing with eyeglasses make IPD measurements to ensure proper
fit. Unfortunately, for any screen viewed simultaneously by more than
one viewer, it's likely that the IPD on screen will not match all viewers.
Next comes screen size. In a movie theater, it is possible to adjust
the IPD for the screen size, either to a desired mean or, perhaps, to a
maximum, so that no viewer needs to diverge (an unnatural eye movement),
no matter how distant the object displayed.
In theory, at least, the same could be done in a TV set. "Knowing" its
screen size, a processor could adjust the 3D accordingly. I am not
aware, however, of any TV set doing so. So, if a broadcaster transmits
one signal that might be displayed on screens ranging from a 1-inch
mobile phone to a 152-inch plasma, the IPD will vary accordingly,
outside the control of the broadcaster.
I agree with Jeroen that the super-giant plasmas are unlikely, and so,
too, are one-inch screens. But a range from 2.5-inch on a mobile device
to 82-inch (Mitsubishi 3D TV offering) seems quite possible. That's
almost a 33:1 range in IPD. Drop it to a 60-inch maximum and a 5-inch
minimum, and you're still looking at a 12:1 range in IPD.
If a broadcaster were to restrict IPD to 41 mm so that children won't
have to diverge (and, even then, SOME children might have to diverge)
and picks even just 60 inches as the largest TV, then, on a 30-inch
screen, infinity will be roughly 20 mm, which, for an adult male with a
65-mm IPD, would be like looking at something very, very close.
Finally, there's viewing distance. Let's assume the roughly 2.5-inch
mean adult IPD. If there is a 2.5-inch negative parallax on screen (the
left-eye image is 2.5 inches to the RIGHT of the right-eye image), then,
for someone with a 2.5-inch IPD, the object with that negative parallax
will appear to be in front of the screen at half the viewing distance.
If someone is watching from Jeroen's 10-foot distance, the object
appears to be five feet in front of the screen -- regardless of screen
size. If someone is watching from a six-foot distance, the same object,
portrayed at the same time on he same screen, appears to be three feet
in front of the screen.
Suppose the object is a mannequin, held sideways in the z-axis (pointing
in and out of the screen), pointing halfway out of the screen and
halfway into it. Given the numbers I used above, a viewer at a six-foot
viewing distance sees a mannequin six feet long. But a viewer at a
ten-foot distance, looking at the same screen at the same time, sees a
mannequin seemingly ten feet long.
Jeroen would like broadcasters to deal with that. I'd like to know how.
TTFN,
Mark
On 3/6/2010 11:46 PM, John Shutt wrote:
Have you ever seen those "3D" images that look at first glance like
just a pattern repeated over the entire page, but if you stare at it
long enough, you eventually can see a 3D image within? I believe
Magic Eye is one type of the images I am describing. They're hard to
see at first, because your eye needs to focus at a close distance, but
your eyes need to converge at a point much further away. It takes a
while for your brain to decouple left eye-right eye convergence with
focal distance.
This may also be the cause of some people's discomfort with 3D
movies. Their eyes are clearly focused at whatever distance they are
from the screen, but the 3D depth may not be set at that eye
convergence angle for that distance.
This is of course exacerbated with small screens viewed at a closer
distance. You may be seated at the same number of picture heights
away from the screen in a large theater and seated at your laptop, but
your eyes are converged and focused at two different angles, giving
your brain different starting points to "decode" the 3D being viewed.
Is this the difficulty you speak of when trying to create one 3D
program to be viewed on vastly different screen?
John
----- Original Message -----
*From:* Mark Schubin <mailto:tvmark@xxxxxxxxxxxxx>
So any complaints should
go to the broadcaster of the 3D program, or the producer of the
DVD/Blu-ray
disc. They should quickly learn to produce with the optimum
amount of depth.
A tricky proposition when their broadcasts may be seen on screens
ranging from one-inch to 152 inches and at a very broad range of
viewing distances.
Other related posts:
