Mailinglist Archives:
Infrared
Panorama
Photo-3D
Tech-3D
Sell-3D
MF3D
|
|
Notice |
This mailinglist archive is frozen since May 2001, i.e. it will stay online but will not be updated.
|
|
P3D Re: Converting 2D into Simulated 3D
- From: Peter Davis <pd@xxxxxxxxxxxxx>
- Subject: P3D Re: Converting 2D into Simulated 3D
- Date: Fri, 21 Jan 2000 03:49:44 -0700
At 05:38 AM 1/21/00 , you wrote:
>Peter Davis wrote:
> > Or are you talking about some technique for taking a single photo from a
> > single vantage point and, using a computer and some image processing
> > software, generating a second image? If that's the case, it's been done,
> > but I think it's unreliable and very labor intensive. It generally
> > involves cutting portions of the image and shifting them slightly to the
> > right or left to create the parallax.
> >
> > It might be possible to do this automatically at some point with a
> grab-bag
> > of techniques from artificial intelligence, image processing, etc. A
> > computer which "recognizes" people and objects in a photo might be able to
> > reconstruct a model of those objects from its own database, and then
> create
> > a second image of that model from a slightly different view point. I
> don't
> > think there's anything marketable yet that can do this.
> >
>
>This has been possible for several years. I saw a paper at SIGGRAPH a
>while ago about reconstructing 3D computer models based on a single
>photograph of a scene and some educated guesses about various
>dimensions and knowledge of the optics used to take the photograph. I
>think they may have also used knowledge about the time of day and
>shadow positions. You can probably find the paper in the conferendce
>proceedings from the mid-1990's.
There's a big difference between a SIGGRAPH paper and a marketable
product. The paper usually means a couple of researchers have managed to
accomplish this for a few, generally carefully chosen, special cases.
I know some folks at the MIT Media Lab were doing similar experiments with
a "range" camera ... a single camera that measures distance to
surfaces. They were able to reconstruct 3D models of objects from the
range information and then rotate the models. But guess what! When you
rotated the model so that part of the back surface should have been visible
... there was no back surface. All the objects were half shells of
objects, with no back or interior.
-pd
--------
Peter Davis
Funny stuff at http://people.ne.mediaone.net/peter_davis
"The artwork formerly shown as 'prints'."
Resources for children's writers & illustrators:
http://people.ne.mediaone.net/peter_davis/cwrl.html
|