Has 3D Video Finally Arrived?

pospisil writes to point us to an enthusiastic writeup on Tech.blorge.com about a 3D display technology just launched at the eGames Expo in Melbourne. The technology, from a company called Fountain Consulting, is set to ship in January. From the article: "The Vortex Home Entertainment System isn't just set to revolutionize 3D forever, they have revolutionized it. With a library of 500 current PC-based games titles converted to flawless 3D, and even the ability to convert 2D live television into 3D live television, as well as pre-recorded movies on DVD, Blu-ray and HD-DVD." There is no second source for this story. Exciting news if it pans out.

If they claimed it for games only it might be real

[Animats]

It's straightforward to do this for 3D games, because the system has real depth information. Just use shutter glasses [ray3d.com] and render alternate frames with the viewpoint shifted by one eye separation distance. That's easy, and looks good if the system can render upwards of 70 fps.

But any scheme for converting existing 2D content to "3D" will probably fall somewhere between "looks stupid" and "generates splitting headaches".

Stereo vision doesn't do anything useful for objects more than a few meters away. It's most useful for close work, which is rare in games. It's more useful for mechanical CAD, medical imaging data, and similar stuff you need to view close up. Which is why 3D movies, TV, games, etc. never really caught on.

flicker++

[lucas teh geek]

haha americans, for once I can say I was there... and it sucked. perhaps it'd be useful if you want to induce epileptic fits in people, but honestly, the "3D Video" stall was really lame. people would pick up the glasses, look for a while wondering why nothing happened, then someone would point out then needed to be infront of the little sensor/projector thing. so they'd move in front of it, and think to themselves "wow, this looks like crap and is giving me a headache" and put the glasses back down. total turnaround time of about a minute.

Re:My willingness to suspend disbelief...

[Joe Random]

I should hope the GP doesn't lose all depth perception when he closes one eye

Humans compensate for this in several ways. One is to move the head slightly side-to-side. This parallax information can help you in determining distance. Here's an example [sunpig.com]. Note that this requires extra information, so it's not useful in the case of upsampling 2D to 3D.

Another method is to notice when one object occludes another. That could possibly be automated, but you'd need some very sophisticated image recognition and tracking technology. Possibly some pre-processing, too, to avoid objects suddenly "jumping" along the z axis as their size changes force their calculated distances to be modified.

There's also the fact that we tend to know the relative sizes of various common objects, and comparing that to their perceived sizes can give rough distance information. That would require image recognition technology of a degree that we don't currently have, though.

So it looks like occlusion is probably the only method that could glean 3D info from a 2D source with any degree of accuracy, and I can't imagine that that's be very accurate or, indeed, always possible. Plus, I suspect the results would look like a pop-up book, with different portions of the image represented as flat objects on different planes rather than 3D objects.

Re:My willingness to suspend disbelief...

[Anonymous Coward]

> Theoretically, on a scene where the camera is panning, you could have one eye see one or two frames
> ahead of the other. It would look like both eyes seeing it from different perspectives.

This does work, as long as the camera is panning perpendicular to the scene, and nothing in the scene is moving too much over those few frames.

I know a guy who made a stero pair this way, from video of Celine Dion singing that Titanic song. They said it couldn't be done, but he proved them wrong! Hahaha!

Re:Sure..

[x2A]

It's not that far fetched. There are all sorts of cues you can use to tell depth and then you can extrapolate an "off by 5 degrees" image; slightly stretch some pixels and shrink others for one eye, and the other way round for the other eye, will make one eye seem to see more of one side, and the other see more of the other.

http://www.youtube.com/watch?v=zGEQELp0uqA [youtube.com] (jump to around 2:30 to see example)

http://www.youtube.com/watch?v=VuoljANz4EA [youtube.com] (more examples)

And these use a single picture to work from. If you have a moving picture, you have even more information to work with. There may be small artifacts of cause, but these could easily blend out when applying to a moving picture.

It is all in the brain....

[aepervius]

If I remmember correctly, if the brain is used to the item you are looking at, even looking with one eye you will ahve the eprception of depth, NOT because one eye is enough, but because the rbain "remember" it. Now try the same (only 1 eye) in an unfamiliar item or environment, that the brain cannot recognize.

Re:Sure..

[Anonymous Coward]

Listen, if you do not know what you're talking about, then why don't you simply refrain from commenting? It's quite baffling to see a big part of the /. crowd go "It's impossible to get 3D info from a 2D stream!". Go educate yourself. There are algorithms available that are able to do some pretty good guessing as far as extracting 3D from 2D images is concerned. It is based on a combination of motion vectors, parallax recognition, scaling effects and luminance. Sure, it doesn't work all the time (we're still dealing with 2D in essence), but some of these algorithms deliver some pretty impressive 3D streams (and having worked with these algorithms myself, I actually DO know what I'm talking about).