DVD Player Displays 2D Movies in 3D

Anonymous Writer writes "A company called Dynamic Digital Depth that wants to bring 3D television and movies to the mainstream claims to have developed a system that allows you to watch current 2D DVDs in 3D. They claim the TriDef DVD Player uses image analysis methods, developed by the company for their 3D content conversion service, to convert 2D video to 3D in real-time based on 3D depth cues in the original movie. It is the same company that produced the TriDef Movie Player software for the Sharp Actius R3D3 autostereo display notebook. "

I remember this...

[iapetus]

I remember reading about converting 2D movies to 3D when I was at primary school. Since then I've been through secondary school, two university courses and two jobs, and I'm still no closer to being able to watch the things.

Still, I look forward to being able to read ten years down the line about an amazing new device that can display current 2D movies in 3D.

Independent review?

[Joseph Vigneau]

Has there been an independent review on this technology? I notice all of the links in the story point at the vendor's web site. Until then, call me a skeptic

Or is this just an ad story?

More power to them!

[GillBates0]

I've always been a 3D freak - I've played Anaglyph 3-d Quake [iprimus.com.au], I collect stereograms [eyetricks.com] and routinely watch documentaries in 3-D IMAX [neaq.org].

I'm also firmly believe that VR and 3D displays are the Next Big Thing (TM) - atleast I hope it is. So I say more power to Sharp, DDD and other folks who're trying to make my dream a reality.

On the other hand, I'm not convinced by their "image analysis" based on depth cues:

hey claim the TriDef DVD Player uses image analysis methods, developed by the company for their 3D content conversion service, to convert 2D video to 3D in real-time based on 3D depth cues in the original movie.

As far as I can see converting current 2D media to 3D would require a great deal of human intervention - there's only so much that you can glean from image analysis (possibly hidden edges, object sizes and other CG cues). The bottom line is that it would take a human to tell if which of the two objects on the screen are supposed to be closer to the viewer. That alone IMHO would kill any efforts to bring this to the mainstream media business - it would be more fruitful to focus on cheaper/better techniques to create new 3D media.

Re:Dubious

[bestguruever]

I won't believe it until I see it either, but it does seem possible to a limitted degree. What I imagine this as is using stereo seperation to enhance the existing depth cues.

Meh

[JMZero]

If 3d was impressive enough to sell lots of units, they'd still be making lots of 3d movies. They aren't, because the technology for displaying 3d is still not impressive enough nor widely spread. Once there is truly impressive 3d displays that can be widely deployed, the content will come naturally.

As an aside, I'd love to see Pixar render out a version of Finding Nemo for IMAX 3D - I think it'd be amazing, and would be a relatively small cost. If it was a success, they could do their whole catalog.

Re:Dubious

[sinrakin]

The principle seems straight forward enough. You don't have enough 3D info in a single frame, but you have lots of frames. So as objects move, or the camera pans, you can tell by their apparent positional shift how far from the observer they are. Assuming the software can recognize and track some basic objects, it can make reasonable inferences about their depth into the scene. How it then displays the depth is another issue.

Re:Dubious

[moviepig.com]

There simply isn't enough information in a 2D image to construct a 3D image.

There's plenty of info to construct a 3D-image. There's just not enough to construct the 3D-image.

Part of the bizplan likely involves consumers not caring.

Enhanced for stereo, colorization...

[dpbsmith]

OK, I'm curious to see it--but I don't believe a word of it. My brain is capable of converting 2D presentations into 3D using depth cues. I suspect my brain is better at it than their software is. And that wherever their software falls short, there will be an intense mental irritation factor.

In the fifties, a sound engineer whose name escapes me devoted a _lot_ of effort to applying electronic filtering to add a stereo effect to Toscanini's recordings, with the idea that he was preserving them for posterity. Toscanini's recordings and reputation have survived, but it's noteworthy that all the CD remasterings are in mono.

I don't think I've seen any upsurge of interest in "colorized" black-and-white movies, either.

I would expect automatic 3D to suffer from the same issues as colorizing: problems at the edges where things are entering the frame, problems with things that are in the background and hence out of focus, scenes that consists of thousands of moving objects (crowds, tree leaves flexing in the wind, sunlight glancing off rippling water) where the cues are imprecise and the computational effort needed to track thousands of objects is intense...

Re:Dubious

[varaani]

Not really that reasonable. If you look at the results of current optical flow and disparity estimation algorithms, they're really not that great. Discontinuities of the image (edges) are a huge problem, as is the whole top-down/bottom-up/gestalt-ordeal, and these have not been solved in any satisfactory manner.

To reconstruct the 3D scene generating the 2D images is effectively to solve vision, in its entirety. In real time, no less. So I would guess that they're doing something quite simple. I'd love to see it, but the information on the site is quite scarce. I'm just hoping that someone is not manually pulling the strings behind the scenes.

Re:Dubious

[Anonymous Coward]

There is quite an assortment of depth cues, and only one of them is missing in still frames of 2D movies: stereo-parallax. Motion parallax, depth of focus, lighting and perspective shortening are still there and your brain does a good job of reconstructing the scene just from these. The problem is that your brain also very intuitively recognizes that the scene isn't real because of the missing stereo parallax. If you close one eye, it becomes much harder to immediately tell whether you're looking at a picture or the real thing (given that the scene/image is far enough away so that focus doesn't matter). The goal is to recreate the missing depth cue from the remaining cues.

Re:what does it add?

[swerk]

At first, motion pictures themselves were a novelty. Synchronized sound, color, wider viewing area, hell, even cut perspective changes and camera motion -- all these were novelties. They added another dimension (sorry, sorry!) to film, enhancing the experience and growing to lose their novelty status. Rather than films that served only to show "look, you hear my voice while you see my lips move!" or "look, bright colors everywhere!" those things just became standard filmmaking tools.

Time will tell whether 3D movies are viewer-friendly and/or affordable enough to really catch on. The special-glasses approaches have been too gimmicky or glitchy, perhaps this kind of display will get it right.

And this latest step, analyzing motion cues and faking a 3D movie out of a 2D one, well even if it sort-of works, it's a pretty cool idea. If 3D displays become standard/expected equipment, we'll still be able to play our old movies without having them look completely outdated. Hopefully the "original format" option will still be there, for us anti-colorization, anti-pan&scan folks. :^)

Even if it is possible...

[spaeschke]

I still don't think it's a good idea. Much like colorizing black & white movies, this is changing a movie beyond what the original director ever intended. Even if you're not a purist about this sort of thing, the results would probably still be lousy because it was never in the directors mind in the first place.

As with colorization

[lildogie]

Directors will complain that 3D-izing violates their artistic integrity.

"My movie was written and directed for the flat screen!"

yada yada yada