Become a fan of Slashdot on Facebook


Forgot your password?

Disney Research Creates Megastereo - Panoramas With Depth 48

mikejuk writes "Disney Research has made a breakthrough in implementing the technique of acquiring depth information from a simple camera scan of a scene. For a perfect panorama you need to rotate the camera around its optical center, i.e. just rotate the camera. However, if you just rotate the camera about itself you don't get any parallax effects — which is why it makes the stitching together easier. If you want to get 3D information from the sequence of shots you need parallax. This means rotating the camera mounted on an offset arm or just moving the camera along an arc in your outstretched hand. The big problem with this method is that the parallax makes it more difficult to fit the mosaic together, and this is the problem that the research team has been working on. Using a range of different scanning methods the results can be converted into high resolution panoramas automatically complete with 3D information."

This discussion has been archived. No new comments can be posted.

Disney Research Creates Megastereo - Panoramas With Depth

Comments Filter:
  • Nice. Given Disney's interest in converting 2D movies to fake 3D movies, it's not surprising that they're investing in this. It looks like they're able to do this without having to create imagery to fill in occluded areas, which is a problem with some of the other approaches.

    Next, Google StreetView needs to be redone with this level of quality.

    • by Romwell ( 873455 )
      Google StreetView already has a 3D mode, though, although its output is not that controllable (probably because it has been preprocessed into anaglyph - a pity for those with 3D monitors). The ability to adjust separtion/convergence would be a simple add-on, though. Also, Disney's panoramas are closer to real 3D than just a stereo-pair based imagery; there is one degree of freedom in viewing positions - and that's all you really need for most applications.
    • You can't use this technique to convert and already filmed 2D to fake 3D. This technique works by filming those "occluded areas" already, much like a 3d camera would.
  • by StripedCow ( 776465 ) on Sunday June 23, 2013 @03:32PM (#44086993)

    Are there any algorithms out there that can take a movie, and produce a sharp photo (or a series of sharp photos)?
    By "sharp", I mean much sharper than each of the individual frames of the movie.

    I.e., the algorithm should use information between subsequent frames to sharpen the image.

    • Re: (Score:3, Informative)

      by Anonymous Coward

      Yes there are. See for methods and limitations. They are not yet in movie-scale but can convert a sequence of images from a single rigid target into a more detailed one if sub-pixel accurate registration is done between images.

    • Yes; search for "super-resolution []".

    • Re:Curious (Score:4, Funny)

      by I'm New Around Here ( 1154723 ) on Sunday June 23, 2013 @04:34PM (#44087321)

      Sure, I see Abby do that on NCIS all the time.

      It works best if it's a reflection on a pair of sunglasses, though.

      • You have to have someone else stand over your shoulder and say "enhance image" otherwise it won't work.
      • by chinton ( 151403 )
        Various CSI teams from Las Vegas to Miami have been using these algorithms years before they were adopted by the Navy. Although, the civilian applications seemed to be stronger at inferring pixels, thus allowing a near-infinite zoom.
    • Re:Curious (Score:4, Interesting)

      by drkim ( 1559875 ) on Sunday June 23, 2013 @11:48PM (#44089251)

      Are there any algorithms out there that can take a movie, and produce a sharp photo

      Of course. And it's free:
      RegiStax []

      It's mostly used by astronomers, but works quite well for any series of images. Of course, it works on repeating frames. If things are moving, them they aren't 'repeating' and won't be processed correctly.

  • The author has a weak grasp on how stitching software is affected by parallax. Also, it would still be technically easier to to with two cameras or a stereo camera.
    • by Romwell ( 873455 )

      Also, it would still be technically easier to to with two cameras or a stereo camera.

      Yes, it will be, but there is a serious lack of 3D stereo equipment on the market today. Affordable stereo cameras are not of high quality, and I have yet to see one with adjustable lens separation. Two camera solutions come at twice the cost, size, weight and battery worries.

      Also, if I can't afford one D3x, I surely can't afford two of those plus a sync kit and a tripod, nor would I want to carry such a setup with me on a hike. There are plenty situations when all you have is one camera, and you want to

      • by Anonymous Coward

        Their algorithm works best with images taken no further than 4 degrees apart. They have examples with 8 degrees between the images, but it starts to fall apart at that point. With 4 degrees between successive images, you'd have to have 90 images for a full 360 panorama. Unless you've been filming your panoramas, you've probably only been taking about 12 images full circle, and not in the right geometry for their algorithm either.

  • by Romwell ( 873455 ) on Sunday June 23, 2013 @03:58PM (#44087133)

    I've been a long-time panoramic and 3D photography enthusiast, and have gigabytes of data that could be fed straight into this software.

    Shooting a 3D photograph is easy (just take two frames, correct issues in software later); shooting panoramas is easy (let the stitchers do their job); shooting a 3D panorama has always been too much work for me.

    It's a pity I can't get my hands on any working code yet, and any commercial product is probably way off in the future.

    Also, here's a link to their paper [][PDF] for those interested. It's quite readable.

  • Very old news (Score:5, Insightful)

    by Begemot ( 38841 ) on Sunday June 23, 2013 @04:06PM (#44087169)

    It's based on this invention:

    Yael Pritch (who was involved in this project in Disney) was also involved in this research in the Hebrew University.

    So it's a breakthrough, but a very old one and, more importantly, somebody else's

    • by Romwell ( 873455 )

      Thank you for the link! It's old news for you, but some of us are not so up to date.

      It seems like some of this research might have been implemented in this commercial product. I wonder if it is so.

      • by Begemot ( 38841 )

        Sure, pleasure. Re commercial products. I found these guys: and Not sure how successful they are. The app seems to be struggling (probably heavy duty calculations are not ready for mobile yet). I simply looked up what Yael Pritch did in the past, apparently she was the founder of this company. Also check out this: Apparently Sony have infringed this patent.

        • by Romwell ( 873455 )
          I have used Snapily3d services as of this year - so they are active at least - and was happy with their service. Didn't know about HumanEyes until today, though.
    • Cool of you for finding the original link! I actually remember this being exhibited at the HUJI Open Day fair in 1999, and (as a prospective student) had a chat with Yael who explained a bit about how it worked. I'm happy to see that this little project has progressed.

      Your criticism however seems out of place; if Yael is a Disney Research employee then this is a Disney Research invention, it is not "somebody else's". I'm quite sure that the team has not been sitting idly these past 14 years, and that the
      • by Begemot ( 38841 )

        I do not criticize Yael, she's super cool. I do criticize the language of TFA that made it look like a new and original work, and not an improvement over a 15 year old invention, whose original inventors are Prof. Shmuel Peleg and Bob Rosenschein (the former CEO of Yael's current paper expands on it in details.

    • Which points to the shocking state of the patent system, where 'omnistereo' can be patented in 2011 [] despite having been openly published years earlier. Blimey, even stereoscopic movies were still being patented in 2005 []! (oh yes, that 'rectilinear' makes all the difference...)

    • Disney linked to recycling very old tales? 3D Pics, or it didn't happen!
  • Isn't this similar to the 3D Sweep Panorama feature that has been available in some Sony cameras and smartphones since 2010? []

  • Disney Creates? (Score:4, Informative)

    by onyxruby ( 118189 ) <`ten.tsacmoc' `ta' `yburxyno'> on Sunday June 23, 2013 @05:14PM (#44087489)

    Anytime someone reads Disney creates they should substitute the word "copies" until proven otherwise. Disney has a long and storied history of intellectual property theft when it was in their best interest. They are arguably the greatest hypocrites in the world about IP, even more so than Hollywood themselves. They are always the ones that hold the most radical of views in the MPAA and are very quick to hold condemn anyone else and take away their rights. They have also been stealing from the public domain and other individuals for ideas for decades. A quick Google search can find example upon example of their bad behavior.

    • Re: (Score:2, Informative)

      by Anonymous Coward

      Disclaimer: I used to work for Disney Research.

      This may be true of other parts of Disney, but Disney Research is only a few years old, is managed by researchers, operates pretty much independently like an academic institution (though with the goal of developing tools the rest of Disney can use) and everything I've seen has been to the highest standards of scholarship. Take a look at the SIGGRAPH publications here: From the last few SIGGRAPHs and SIGGRAPH Asias almost 10

  • I repeat: cross-eye is the only stereogram that does not require additional equipment AND allows normal 2D view if you do not feel like crossing your eye.

    Why protein structure journals got it for at least 20 years now and general purpose 3D photography cannot get it?

  • Disney researchers in Guangdong Ni Hao Liaoning Guangxi Zhuang performed the research.

    It's logo is Mickey Ni Hao Mouse

Happiness is twin floppies.