[TheZenPsycho]

Actually, the convergence problem is theoretically solvable with a light-field projection system. That is, instead of having 2 images that are simply "fed" into each eye but projected onto a flat screen that is not converged with the 3d disparity "information" contained in the images, you have a system where the screen emits a light field that your eyes can focus on at any depth. We already have consumer level light field cameras and attachments. the displays are tricker, but possible, traditionally through the use of hexagonal lenticular lenslets carefully registered onto a 2d "flattened" projection of the light field capture. This is the "True hologram" dream mentioned. The primary limitations being the level of precision and resolution with which you can register the lenslets with an image.

that's fine for a display but is obviously impractical for a large scale theater projection system. Just an arbitrary possibility I thought of just now: suppose the projection screen had a retroreflective surface- That is, light projected at the screen gets returned at exactly the angle it arrived at. Combine this with a domed mirror and a backwards pointing projector or set of projectors, with all the requisite optics math and geometry work, it may just be possible to project a lightfield at a screen that bounces back at the audience and appears as a tangible hologram to them.

[TheZenPsycho]

I want to be clear that not any of this is easy. I think as soon as you try it, your first attempts are going to be very blurry, squint inducing, and especially very dark. A lightfield display, in order to achieve the same level of perceived brightness as a traditional 2D display, needs to generate 2-10 times (perhaps much more) more actual light, with all the requisite power requirements that entails, since the light is distributed directionally instead of diffusely.

Going in this direction basically takes the level of resolution and precision in image reproduction we've achieved back a decade or 5, since the "pixels" or resolution units are spread over many more views than just 1 or 2, or perhaps something more recognisable as a continuum of "infinite" views, or whatever number is visually indistinguishable from infinity.

[dTal]

Why would it need to generate more light? Suppose you told your lightfield display to simply display all white - wouldn't its light distribution be the same as a white LCD, hence taking the same amount of power?

[TheZenPsycho]

In a projector system you generate light, and then you block it using either photographic film or LCD to produce the image. A lightfield display can be thought of as a 2D, higher resolution generalisation of a lenticular ("no glasses") display, or for example, the display of the 3DS.

In these kinds of displays, you have a light generator (LED or flouro or reflection) with the same surface area as a normal print or display. But to produce the directional light, for say the left eye, you must block the portions of the image that relate to the right eye from the light travelling to the left eye. And vice versa. This halves the amount of light for 2 directional images. thirds it if you want 3. So you end up with an image that is much darker then normal. you must compensate by generating 2 or 3 times the amount of light. This problem gets worse the more "views" you add on. So if you want 10 omni directional views you have to generate enough light for all of them, since, even if you set everything to "white", as you suggest, most of that white is getting blocked from your view.

Or to put another way, you get allocated a smaller source 2d image plane surface area to generate your viewpoint.

[dTal]

Why do you have to block the light? Sure, primitive stereo parallax barriers like the Nintendo DS has do, but nobody uses parallax barrier tech for any proper lightfield application. The system you mentioned, a lens array over a high-PPI screen, blocks nothing; light is only refracted. Ergo, the energy required to run the screen is the same as without the lens array, e.g. a normal screen.

[jameswilsterman]

I was recently at the Exploratorium in SF (Embarcadero). They have a mirror there that tricks your brain into thinking you are seeing an upside-down reflection of yourself hanging in the air about 5ft in front of where the mirror actually is. I actually tried to touch the 'reflection' and my hand just met the air. My first thought was why don't we use these to create 3D images. Would your domed mirror proposal work similarly?

http://www.exploratorium.edu/visit/central-gallery/giant-mir...

[TheZenPsycho]

well yep, but that one has the advantage of using the lightfield already reflecting off of you and simply redirecting it back with relatively simple, shall I say, macro-optics.

The kind of system I'm talking about needs a 2D image with enough "pixels" to fill a volume convincingly instead of just a 2D plane, and the optics would need to be far more complicated, precise, and created at a very small scale which would appear as a textured surface like a fresnel lense or one of those 3D lenticular stickers you sometimes see. Then once you've figured that out, you need to get the projection and the optics to line up precisely- unless you can figure out a way to build in some tolerance to the alignment of the projection.

Aside from that, the goal is essentially the same. To produce a field of light coming out of some "window" with the same directional qualities as the light coming out of a real window.

[Geee]

One solution for holographic lightfields is to use nanoantenna arrays. Multiple antennas in an array, in which each antennas phase and amplitude is modulated is able to produce these directional lightfields by interference, and by modulation it's possible to change the shape of the radiation pattern without touching the antennas. In nanoscale, it could create waves in optical wavelengths.

[maaku]

This is why I still get my news commentary from HN. Thank you.

[repsilat]

It's also (mostly) solvable with active lenses and eye-tracking. What you can do is infer distance from vergence and then adjust the effective focal distance of the screen.

I don't know how cumbersome the lenses would have to be, or how insanely fast the system would need to operate, but it's at least possible in theory.

[tgb]

"retrofelctive" - thank you for using this word! I basically froze up for a good while one time looking at a street sign at night, realizing how it's 'reflection' was different from typical mirror reflection. Now I know what the term for that is and how to build one!

[dTal]

If the screen were retroreflective, wouldn't all projected light end up back in the projector by definition?

[calroc]

Is "retroreflective" anything like a phase conjugate mirror?