Glasses-Free 3D Discussion Televisions / Cinema / Gaming - Blu-ray Forum

globalimages · 09-16-2020, 04:13 PM

Sorry to say, but just don't bother with glasses-free 3D in the sense we all want: a cheap-enough "device" similar to a (TV or projection) screen in which a group of people (without too much constrain) can simultaneously watch a "natural" (i.e. not only CGI) moving scene in "near perfect" stereoscopic glory (one separate view per eye corresponding to aprox. say 7cm of inter-axial distance).

3D film technology has existed for about 100 years and this "problem" hasn't been solved because, well, it "can't be solved".

Or rather, the only way to solve it is:

A) Either somehow track where all the eyes of all the people watching are with respect to the "screen" and somehow shoot the appropriate view to each one (still some could overlap, like someone sitting in a sweetspot behind another person with the "wrong" eye in that area)

or

B) Use any of different methods to fill a certain viewing angle in front of the "screen" with multiple views appropriately separated horizontally so people within this area get the corresponding image-per-eye. Well, for this system, you need basically a view per each "aprox" 6cm of space (the space between the eyes) of the end of the area "cone" you want to fill for different viewers, that is, you need many, many, many views, the more the better (to a point). Thus you have to capture, edit, transmit, etc all those views, or use powerful processing to try to do "fake 3D" and interpolate artificially generated extra views between those you captured and those you are displaying.

You want to do this right? You need as many view pairs as (aprox) 6cm of space "slots" you want to fill in the room. Even if you interpolate, that means 3 views better than 2, 4 better than 3 and, well, to be realistic, you better use at least 5 to 25 views (or equivalent lightfield ray direction information). So the capture "resolution" (number of pixels or views) and bandwidth needed for the transmission of all this information is, well, say 5 to 200 times higher than we are using today. If you think using 2 cameras to capture 3D is a hassle, just ask cinematographers to use the equivalent to an array of say 5 (6, 7 ... 25) cameras, the more the better, or equivalent "distance" holographic (or light field) sensor device (a VERY large "sensor", we are talking >1meter long sensors/array of sensors/array of cameras if we want to do it right).

So yeah, besides pure CGI content (which can generate the necessary large number of views "on the fly" with a powerful enough GPU), natural-looking stereoscopic (different view to each eye) images just can not be shown to an unrestricted group of people spread in the space in front of a "device" without filling such space with as many corresponding views as "slots of eye-pairs" are in such space, be this "pair of images" reconstructed by a lightfield or discretely in vertical slices ...

So in short: it's never going to happen, until it becomes cheap enough to film with a minimum of say "5 cameras" and to store/transmit the information equivalent to "say 5 images" PLUS equip a powerful GPU at the display-end able to extrapolate those (say) 5 views into (say) 20 "fake-but-good-enough" views needed ... And still the field of view of such a display would still be a bit limited and/or a few positions in front of it would not work (easy to fix, just move your head a bit to the left or the right, or a bit closer to/farther from the display).

This is physics, there is no way around it. It's been researched for 100 years since the first 3D film and we still don't have glasses-free 3D displays of any kind (to the "good enough" level that we want). Before you say we have holograms that are close enough, just think of the amount of information a hologram contains (to transmit it per second to form moving images) and that it's also a bit limited in viewing angle, size, etc.

CGI scenes: not too much of a problem, the information needed to transmit is small, powerful renderers could make multiple views at the display end. Natural scenes: well, the more views (pixels) you can capture the better, a few in-between could also be artificially fake-3D rendered, much like the 3D-conversions done today on most 3D films. Compression can be used: just send a "large background image", all necessary "textures" and a depth map and have a processor figure out multiviews (mostly what has been done up til now with lenticular TV displays and similar). The results up until now: still limited in all senses (resolution, viewing angles, crosstalks). It can be done pretty good for CGI and static content though, like on these commercially available displays some 10 years old, but that's not what we are talking here, right?

Best 3D Blu-ray Deals