The Science Behind the InfinitEye's Panoramic Virtual Reality Headset

muterobert writes "The Oculus Rift has competition, and it's incredible. The InfinitEye has a 210 degree field of view (compared with the Oculus Rift's 90) and surrounds your peripheral vision in the game completely. Paul James from RoadToVR goes in-depth with the team behind the new device and finds out how high-FOV Virtual Reality really works. Quoting: 'At the present time, we are using 4 renders, 2 per eye. Left eye renders are centered on left eye, the first render is rotated 90 left and the second looks straight ahead, building two sides of a cube. Right eye renders are centered on its position, the first is rotated 90 degree right and the second looks straight ahead, two sides of another cube. We then process those renders in a final pass, building the distorted image.'"

Comments

Out of date

[Anonymous Coward]

Oh come on, i thought we had got over the whole "strapping-two-screens-to-your-head-like-we-did-in-the-70s" thing

this has far more potential and could actually be mounted into glasses you could wear unlike the screen-strapping tech like Occulus and its clones which haven't progressed much since the 70s (complete with Fresnels lol)

http://tech.slashdot.org/story/13/11/13/1612237/demo-of-prototype-virtual-retinal-head-mounted-display

Latency?

[Scowler]

I RTFA, but didn't see a latency (from sensor to screen redraw) spec. Isn't that supposed to be a pretty important criteria for these devices?

Re:Latency?

[Insomnium]

Yes, among other things. They use 2 renders per screen, so you should be expecting about 16ms more latency than oculus. Also they have double the vertical pixels compared to oculus, so you might expect more latency compared to the oculus screen from computations too. Latency comes down to several factors, rendering latency, screen latency and everything in between. pixel to pixel latency is very much dependant on the screen type. Oculus has said theyre looking into OLED screens that have minimal latency (few ms) compared to lcd screens that have about 16ms or so.

Re: Latency?

[Insomnium]

Both oculus and this one seem to use ~1000hz tracking. so its not an issue, tearing comes from display refresh and it's not about the tracking. In a video i saw some guys at occulus break down the latency and it was something like 2ms for usb, 16ms for game engine/rendering, ~16ms for screen refresh@60fps/hz and 16ms for pixel switching.

Issues...

[Junta]

First, for the apples to apples portion of the discussion, that display technology is 45 degree FoV. Given the article is about a project largely of interest because they were ambitious to 210 degrees which is much higher than the still respectable 90 degrees of Rift, bringing in a 45 degree FoV product into the discussion isn't immediately helpful. Now you *could* be suggesting that the technology could do better if they wanted, but until that's demonstrated it would be risky to assume that. The most optimistic reference material I could find about that sort of design said '100 degree FoV could be possible' based on designs that acheived 60 degree FoV' (but that's not exactly an apt comparison, since that material predates DLP which means it isn't quite talking about the avegant solution). In short, Avegant is aiming squarely at private consumption of video content rather than immersion.

Second, a significant driver for these new projects is a realization that HMD isn't a market that can drive a lot of custom, one-off design work right now. In order to get to a technology that people can actually *get* at an approachable price, they are working to leverage mass-market display technologies that are largely paid for by their use in tablet and similar form factor applications. DLP into the eye is a bit more custom and will probably not be as cheap.

Also, this discussion is solely about the display technology, but a very large part of the work that Oculus is focused on is motion tracking, which is a pretty critical component.

Finally, at least that prototype doesn't exactly look like the poster child of 'glasses you could wear', it's still pretty bulky.

I'm not saying that Avegant should pack up and go home, it could be very promising, but that's not a good reason to tell Oculus and InfiniEye that they are on a dead end path either. Avegant doesn't waste available resolution like the alternatives, but currently there is no solution that leverages the full resolution of the utilized technology while also providing an immersive FoV, but the former point might be moot if the tablet manufacturers continue their one-upmanship to the tune of 3840x2160 7" displays.

Single FOV number no longer adequate

[CityZen]

With these somewhat asymmetric FOVs, a single number doesn't provide enough information to understand what you're getting.
What's needed now is the "inside angle" and the "outside angle", where:
- inside angle = how much either eye can see toward the other eye
- outside angle = how much either eye can see away from the other eye
(in either case, measure the angle from "straight ahead" over to the cut-off point where you can no longer see anything)
In a symmetric system, both of these numbers are the same (or pretty close, anyway); you'd just add the two to get regular FOV.
You don't want the inside angle too small, or else you'll feel like you've got a huge nose (or your hand between your eyes).
Making the inside angle large is complicated by the fact that the displays will run into each other.
Making the outside angle large is easy by comparison.

Curved display

[DrYak]

Curved display will be the future, yes. The *far* future.
As said by other in this thread, Occulus and InfinitEye project try to make the technology cheap affordable and mass produced. Thus they concentrate on the cheapest and most available hardware.
Currently, 5" to 7" *flat* displays are ubiquitous in tablets, and thus are very easy to source for Occulus and IntinitEye (and thanks to the "Retina" fad launched by apple, these screens have high resolutions too, thus looks nicely even if blown up to the full field of view, unlike the 200x200 pixel HMD of the 90s).
Currently, curved displays are still something new, sold at a premium for a few "almost-concept" models of LG and Samsung (not even Apple, yet). They are not cheaply available in big quantities. For now.
If the "curved screen" fad catches with smartphone makers, you can bet that the available quantities will go up and the price will get lower. Thus, eventually future iterations of Occulus, InifitEye and other such technologies will be able to afford such technologies and stay within the target price, while getting less bulky and better quality.

As of the lens system: Never mind. It's a software problem. Current prototype of InfinitEye and Occulus *already* have a sub-optimal lens system. But that's part of the design. Instead of engineering a complex piece of optical apparatus that can perfectly and distortion-free project an image into a rectangular field of view, they go for the cheap and affordable len (InifintEye is using even Fresnel lens), and pre-distort the image using shader before sending it to the display.
Again, it's a lower quality solution, but helps avoiding that more than 600$ are thrown in the optical components only.

Also, curved display could help reduce some of the distrotion, if they manage to become affordable in a few years.

Eye focusing is for short distance *only*.

[DrYak]

The "accommodation reflex", i.e.: "how much your eye need to focus" is only used for depth perception at very short distance.

e.g.: If you read a book, are busy hacking/reparing stuff on your bench, etc. your brain notice that the object are near, not only because it needs to cross the eye a lot to focus on the short distance, but also because you eye's len need to contract a lot to accomodate (well at least for those who are of a still yound enough age to be able to do it. Older geeks with presbyopia [= rigid un-compressible eye's len due to old age] just take their reading glasses to compensate their for the same effect).
That's why also, people working on micro-mechanics (clock makers, for exemple) wear special glasses and similar eyepiece that blow up the proportion on what they are working. Otherwise their eyes would have to focus on a way-too-short distance (due to the small scale of their work) and either tire their eyes (eyes haven't evolved very well for looking a sub-milimeter object a few cm eye from the eyes) or not possible at all (such hyper focusing is only possible for children and youngs).

2nd e.g.: That's why "The Magic Eye" type of stereograms feels weird (and some people aren't able to see them) - the parallax of the random-dots evokes distance, but your eyes are still focused at reading distance, and the brain feels that discrepancy a little bit weird. (You need to take the habbit to cross your eyes and focus their lens at 2 different distance). (Which is something that near-/far-sighted people are used to anyway).

The brain doesn't give a fuck about the eye's len focus for anything farther than reading distance. (The len doesn't focus noticeably much, anyway)
Whether the eye len focus on the personne a few meter away, or on the mountain in the distance, it's all the same for the brain. Only the parallax matters (for the person your eye are still crossed toward a few meters, whereas for the mountain the eyes are almost-parallel).

Virtual reality is all about immersing *inside an environment*, not about having sit down and read a book (that's imagination, it's a different form of escapism ;-) ).
Nearly everything happens at a distance (virtually).
Thus the simplest thing is to build the optical part of the display in such way that that this image is focused at infinity (like anything long distance in the real life) and the only 3D effect comes from the parallax (the differences between the screens for each eyes).
It will be completely unnoticeable for 99% of the actions hapening. It will only get noticed on object at very short distance (the occasional door-switch, picked-up document or machine interface). And even then it won't feel any weirder than using reading glasses.

The only *really* weird effect, as mentioned by the AC next to me is the Depth-of-Field.
Under badly lit and dark environment, they eye open the pupil and that reduces the depth of the field (i.e.: the range between the shortest and the farthest point on which the eye is focused. Any object nearer or farther than this range appears blurry. You can't look an object something and still see clearly another one).
As the len of a virtual head-set is perfectly focused(*) on the screen, that means that the whole image is always perfectly sharp, as if it was in a brightly lit day.

As mentionned by the AC, some DOF blurring-effect can slightly make this more realistic.

(*): well for classical old-school expensive almost-custom-made HMDs, the complex optical system is perfectly focused on the screen.
For modern cheap off-the-shelf HMDs, there a plain simple len (of fresnel in TFA) more-or-less focused on the screen. The image is sharp, specially in the center, but the whole image is completely distorted in a "cushion" way. But that's easy to compensate in software by pre-distorting the image in the opposite way(**) and at least, this way the hardware doen't cost 1000$ only for the optical part itself.
But in both cases, the end-effect is the same: you look into the eye-piece and see a sharp picture hanging at infinity.

(**): That's what I'm arguing about curved display: they'll bring a bit less distortions.