At the CML meetup this evening it was said more than once “If a bomb went off in this room, there wouldn’t be any 3D, there simply wouldn’t be any skilled stereographers left!”
Without skilled stereographers, aspiring newbees would need to depend on the plug-n-play, run n’ gun, integrated stereo acquisition devices like this Sony camcorder
See that little knob? It is on the lower right behind the lens casing. That’s the convergence / iris control. I don’t claim to be a stereographer, although I am capable of spelling 3D (a key requirement). In any case, my proclivity is to have easier access to the convergence knob, in fact I’d prefer it if the convergence control was located right next to the zoom control, manipulated by some sort of smooth convergence slider.
Maybe the convergence control is tucked away to dissuade people from getting trigger happy.
Look at these mini 1080p cameras developed by Fraunhofer.
Below you will see a “Through / Under” Beam Splitter 3D (BS3D) rig by Element Technica. Having a camera below facing up allows for more vertical clearance, while limiting the downward rotational angle of the rig.
SterGen has achieved realtime 2D to 3D conversion of generally acceptable quality for the sports arena.
The geometry of the soccer field is known, the lines in the grass, the ball, and other elements.
Maybe some sort of Baysian logic or “Bag of Words” techniques are used to generate parallax values from the 2D image. The field is static and 80% the majority of the background so in-filling after image segmentation looks natural. In fact the details of this system are scant. The above are just guesses.
Sports is such a large market. Seems attractive to run some tests of this system, modified for American football.
Jacqueline, whom you might have met at my previous CES blog, returned to her 3D roots, representing the very stylish Marchon 3D eyewear.
Michael Starks from 3DTV Corp is very inquisitive. In fact he is relentless! Together we extrapolated a lot of information about extrapolation. Specifically extrapolating 28 views from a stereoscopic pair using view synthesis. The number 28 is necessary to power the Philips branded Dimenco auto-stereoscopic 28 view display.
Actually the input format “2D+Depth” is really a side by side image. On the left side is the half resolution 2D texture, maybe the left camera, and on the right side is the half resolution depth map.
Depth maps that are hand drawn look just so nice. Depth maps that are taken from CGI Z buffer look great too!
Depth maps that come from live action look very messy. Not ready for prime time.
However, Depth Maps are not required for that display, one could feed the Dimenco a formatted video file at 1920×1080 with the proper pixel interleave and power the television just fine without using 2D+Depth mode.
-Ethan Daniel Schur