Hi All,

I am a newbie in 3D Graphics part of Comp. science.

I am looking for either a library or an open source code which can
help me project objects in the 3D space, in the same way when we see
3d movies where the objects seem to be getting projected and coming
out of the screen.

Thus, rephrasing it, I am looking for a  library that allows me to
create shapes and project them in real 3D space.

Any information shared on this will be helpful for me. May it be
referring me to the group which can help me or a site also.

Please let me know if I need to clarify more on what I am looking for.

Or is it something that I am expecting too much? :-)

Thanks in advance.

Regards

On Tue, 21 Dec 2010 10:02:05 -0800 (PST)
swapan <swapan.s1@gmail.com> wrote:

> Hi All,
> 
> I am a newbie in 3D Graphics part of Comp. science.
> 
> I am looking for either a library or an open source code which can
> help me project objects in the 3D space, in the same way when we see
> 3d movies where the objects seem to be getting projected and coming
> out of the screen.

You mean like turning a 2D image into a stereoscopic one. This is
almost impossible. There are some methods like camera matchmoving to
extract depth information from a video, but those methods are rather
cruse. Also they are problematic whenever it comes to translucent
surfaces.

> Thus, rephrasing it, I am looking for a  library that allows me to
> create shapes and project them in real 3D space.

You create objects in 3D space. But then you'll also need a display
system capable of showing stereoscopic pictures.

Stereoscopy itself is quite a involved topic... I've got a very lengthy
article about it in draft, I think I should finish it.

Just to prove that I know what I'm talking about: "Elephants Dream" is
an (the first) Open Movie created using the open source software
Blender.

I extended Blender with stereoscopic workflow options, like a
stereoscopic camera object, that automatically takes care of setting
all parameters relevant for rendering based on maximum screen space
separation and convergence depth.

This is the result
http://www.youtube.com/watch?v=L-tmaMR1p3w


Wolfgang

First of all, I must say  ......... Awesome reply and much relevant
information. This is much
more than what I had expected to get in the first reply.

> You mean like turning a 2D image into a stereoscopic one. This is
> almost impossible. There are some methods like camera matchmoving to
> extract depth information from a video, but those methods are rather
> cruse.

Okay, it means that  camera matchmoving can help me transform an
object seen on a 2D screen
to be projected in real 3D space (illusion) by use of a projecting
gadget like a 3D TV.

Means that I will get the same experience that I get when I see a 3D
movie in theater using my 3D
goggles. ?????

>Also they are problematic whenever it comes to translucent
> surfaces.

I am sorry . i couldnt get what you mean by "TRANSLUCENT SURFACE" :-(
Is it th surface of he objects as TRANSLUCENT.


> You create objects in 3D space. But then you'll also need a display
> system capable of showing stereoscopic pictures.

Won't 3D TVs help in this regard? please suggest.

> Stereoscopy itself is quite a involved topic... I've got a very lengthy
> article about it in draft, I think I should finish it.

It would be great if you can share that. :-)

>
> Just to prove that I know what I'm talking about: "Elephants Dream" is
> an (the first) Open Movie created using the open source software
> Blender.


Excellent info. I checked the blender.org for a moment. I believe it
helps making 3D objects but they are still
confined in the 2d screen space. Please correct if I am wrong?

I also check the youtube video.Good share. And it is 3d in 2d space.
correct?

So, does matchmoving help in making this 3d in 2d space become 3d  and
getting projected in real 3d space?

Thanks for the support.

Regards,
Swapan

On Tue, 21 Dec 2010 14:00:14 -0800 (PST)
swapan <swapan.s1@gmail.com> wrote:

> Okay, it means that  camera matchmoving can help me transform an
> object seen on a 2D screen
> to be projected in real 3D space (illusion) by use of a projecting
> gadget like a 3D TV.

It can help, but it won't do it fully automated. There are some
algorithms that claim being able doing it, but I highly doubt their
quality. There's just too much information missing
 
> Means that I will get the same experience that I get when I see a 3D
> movie in theater using my 3D
> goggles. ?????

Unfortunately no. It requires a lot of postprocessing to get right.
"Alice In Wonderland" by Tim Burton was upconverted with decent result.
"Clash Of The Titans" also used post-production 3D and the quality got
abysmal.

> >Also they are problematic whenever it comes to translucent
> > surfaces.
> 
> I am sorry . i couldnt get what you mean by "TRANSLUCENT SURFACE" :-(
> Is it th surface of he objects as TRANSLUCENT.

Translucent like glass, especially if the surface is not clean. Any
upconversion algorithm will get in trouble to decide which depth to
place each pixel: The depth of the objects visible through the glass or
the depth of the markings of the glass itself.

> Won't 3D TVs help in this regard? please suggest.

Well yes, as to stereoscopic projection systems or any other
sterescopic display.

> Excellent info. I checked the blender.org for a moment. I believe it
> helps making 3D objects but they are still
> confined in the 2d screen space. Please correct if I am wrong?

The geometry is 3D, however when rendering the geometry gets projected
into a 2D image space. However it is possible to render into a voxel
image, i.e. a 3-dimensional volumetric image. Maybe you already
experienced a medical CAT or MRT scanning. The data generated there is
such a volumetric image.

> I also check the youtube video.Good share. And it is 3d in 2d space.
> correct?

It's a stereoscopic video.

First you must understand that all those "3D" stuff going around at the
moment isn't 3D at all. It's stereoscopy: Two images, one for the left,
one for the right eye superimposed on the same display, with some
separation mechanism so that each of the viewer's eyes sees only the
one image designated for it.

But there is no projection happening at all. Both the images are flat
and reside entirely on the flat screen.

The trick is, that stereoscopic pictures contain small differences in
the horizontal placement, depending on the distance. This is called
"Parallax". A lot of newbies to stereoscopy tend to toe-in the cameras,
but this will cause perspective distortions, trapezoid in the vertical
and logarithmic in the horizontal. The correct way is to apply a lens
shift -- http://en.wikipedia.org/wiki/View_camera#Shift -- keeping the
major optical axes parallel. Or toe-in and de-distort the image
afterwards.

> So, does matchmoving help in making this 3d in 2d space become 3d  and
> getting projected in real 3d space?

It can deliver additional info, but it can't do any magic. There is
additional work required.

In my professional* opinion, being the stereographer of "Elephants
Dream 3D" and also having written all the extra code for it:
Post-production stereofication is not worth the effort. If you want
decent stereoscopy you've to do it all along the workflow. Having done a
complete stereoscopification beginning-to-end I even say, that doing
all in stereo is easier and the workload of post-production
stereofication afterwards. And I'm not alone with this opinion. For
example James Cameron thinks the same; Avatar has been produced in
end-to-end stereoscopic workflow.


Wolfgang

*: Hey my efforts even won the Blender Institute an award:
http://www.3dmedia2010.com/en/3d-film-festival-0