3D SPEX (and LCS in general)

[P3D <URBANIC@xxxxxxxxxxxxxx>]

Andrew,

thanks for your thoughtful insight.  I am glad to see that people are actually
thinking about this stuff instead of just reacting to what vendors promote.

>>John Urbanic <URBANIC@xxxxxxxxxxxxxx> wrote:
>> I wasn't claiming that sync-doubling was the cheapest, just
>> the highest quality.

>Andrew:
>Page Flipping Systems which switch at 100 or 120Hz offers the highest quality
>since they show full resolution in both fields. Sync-doubling offers half
>this "quality".

No question that a full-frame high-frequency system is the most obvious
way to go, as noted in our "3D Theory" pages.  But as you yourself note:

>But, of course, you have to have the appropriate equipment to be able to do 
>page flipping at 1024x768 at 120Hz...

and this plain doesn't exist except in a few very high-end pieces of equipment
that no one wants to pay for.  Why does the rest of the world need 1024x768
at 120Hz?  As I mentioned, we wouldn't have any customers at all if we held
our breaths until the rest of the world saw it our way, and then enough of them
threw out their otherwise good monitors to buy our design.

Also, though page-flipping may seem elegant, even if everyone had the
hardware capability today, Windows doesn't really support it.  Windows NT
can't even fake it.  And don't even think about Digital Video Disk.  We
have a DVD product, and it does provide beautiful full-motion, high-res,
zero-flicker 3D.  If we had a page-flipped system we couldn't do this without
re-writing the operating system and video drivers for every drive on the
market.  With sync-doubling, it was a natural progression from our still
photography product and didn't require the least hardware upgrade for our
customers.  And, we don't have to touch any of their crucial system software
like video drivers or the registry.

>Yes it appears to be true that a 3D half resolution image appears
>to have approximately the same resolution as a 2D full resolution image.
>(I use field-sequential 3D video all the time and am always thankful
>for this principle.) But if you also compare a 3D full resolution image to
>a 2D full resolution image you will certainly notice the difference.

More resolution is always nicer.  The nice thing about sync-doubling is that
you can pick any resolution you want (you can walk into any store these days
and buy a 1800x1600 pixel card and monitor).  If you don't have "enough"
resolution, go higher.  We find that 1024x768 with at least 16 bit makes a
good starting point for photographic material.  We picked that spec because
it worked, not because of some inate limitation in our equipment.  We can
always go higher (and do).

>> DON'T USE LOSSY COMPRESSION FOR 3D.
>I can't help but notice some double standards cropping up here. First you
>say that it's OK to use 3D at half the vertical res. of 2D images but then
>you say that you need higher quality images in 3D than in 2D...

Not at all, although I can see the confusion.  Besides the fact that vertical
resolution is much less important for 3D than horizontal resolution (several
people told me they will have "line-pair" cards at their next club meeting
as they found this fascinating), just selecting a lower resolution is a lot
differant than a lossy compression scheme.  To oversimplify considerably,
JPEG works by "smearing" out the image.  The higher the compression the more
high-frequency information you lose.  High-frequency information is primarily
edges and bounderies in your picture.  This is exactly what you don't want to
lose in 3D, and results in many image quality problems, the most obvious is
known in these circles as "cardboarding".

A second HUGE disadvantage that I didn't mention earlier is that most lossy
compression schemes (and JPEG is one of these) get worse with every iteration
of compression (known as a cumulative loss).  For many applications this is
a non-issue (our Web pages have more than a few JPEGs).  However, in stereo-
scopy it is very important to be able to adjust, or re-register, an image
for differant circumstances.  Every time you adjust a JPEG even a bit, and
then re-JPEG it, you have just smeared the image some more.

>I can say now that I wouldn't be viewing all the wonderful 3d images
>on VREX's www.3dexpo.com site if the images weren't compressed.

That's true.  We, however, made a very conscious decision that if setting
some mimimum standards meant we can't provide "browsing" quality images,
then that was what we would do.  Two years ago that meant we didn't even
intend to offer a Netscape "Helper" application.  This year, we have many
customers trading via the Internet (but not casually browsing; the images
are still at least 500K).  Maybe next year will see .neo files in displayed
form on Web pages.  If not, then the year after.  Our strategy is to do only
high-quality imagery in whatever medium works, not to dumb down the quality
until it fits everywhere.  The rush of technology (things like Digital
Video Disk, and increasing Web bandwidth) seems to be our friend in this
regard.

>As an example, my vote for the best image on the site goes to:
>http://www,3dexpo.com/gallery/gallery_xav/xav_02h_640x480.htm

We, on the other hand, feel that 640x480 is far too low of image quality for
non-novelty viewing by the general public, that the compression to get it into
60K results in serious 3D loss (I personally see a lot of cardboarding), and
would not publish this in an educational CD.  Fortunately, the market assures
that there are always companies to meet the market demand for this
quality/image-size selection, so I can't begrudge you your choice for a
second.  Of all the materials we have licensed, we have downloaded many from
the authors, but never once pulled the published version from their Web page.

>I'd still like to hear more about the .neo format.
>Are there any details available?

You can go directly to the .neo feature page at: http://neotek.com/neo.htm
However, we are not promoting this as any kind of standard.  Indeed it is
proprietary, and we spent a lot of time and money developing something from
scratch targeted directly at stereoscopy.  And then, we used feedback from
our customers and added a lot of functionality from Version 1.0 to Version 2.0,
which seems to be pretty popular (and, yes, we gave free upgrades to 2.0 in
gratitude).  There is very little incentive for us to give this away, and
there are several hidden features that we will announce when appropriate.

What I am suggesting is that you folks consider what the best choice for
an electronic stereo standard should be, and then consider your options.
You may find some of the .neo features useful.  Go ahead and incorporate
them.  Just don't let someone rename an extension from .JPG to .JPS, put
a flag in to state that it is a stereo format and then convince you this
just happens to be right for stereoscopy.  As this and many other debates
on here prove, 3D has been hampered more by compromises over the years that
anything else.  Don't compromise just because someone wanted to published a
"spec" and didn't want to spend more than a day implementing it. If you don't
know that JPS is just JPG, try renaming it and open it as a JPG with any
paint program you have.  A cute trick, but it tells you how much effort to
accomodate stereo really went into it.

John


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