3D is not without problems

[Florin Andrei]

http://www.abc.net.au/unleashed/stories/s2813511.htm

Is this the decade of 3D? It might look that way, but we'd all better hope it turns out quite differently. You see, 3D is not good for you.

How can this be? Isn't the real world in 3D? Yes, the real world of objects is definitely three-dimensional. But that's where the similarity ends. What you're shown on a movie screen - or soon, a television - is not true 3D. That's the source of the problem.

[Jeff Coatney]

For immersive tech like VR and IMAX, which fills your peripheral vision from port to starboard, this is all too true. However, for a home 3DTV environment, where you view the picture from distances that allow you to also see the rest of the room normally, only about 40% your field of view is being tweaked by the parallax effect. This should mitigate much of the binocular disphoria caused by prolonged viewing. But 3D should still be enjoyed in small doses, and contrary to how some professionals shoot their 3D, the intraoccular distance should never be more than 65mm. Yes, the effect of 3D is possible using a wider lens separation, but its about the physiological effects that should discourage it.

[Markus Stone]

It's certainly an interesting article, but I'm not sure we're facing an epidemic of depthblindness. Head tracking VR is quite a different experience to watching something on a screen - there's much more orbital motion of the head involved, and therefore possibilities for just good old motion sickness to arise.

I imagine many people on this forum have seen Avatar - I'd be interested to hear if anyone experienced this effect afterward - but only AFTERWARD mind you. The thrust of the article is saying that when in the real world again, you will have difficulty resolving depth, leading to accidents.

The article doesn't say that nausea is part of the symptoms, btw which could be ascribed to the brain having issues with either bad 3D or the breakdown of accomodation and convergence (particulalrly if it occurred DURING the movie). This would be the exact opposite of the effect he is discussing, ie; the brain rejecting the 3D experience, rather than becoming so immersed it rejects the real world.

So - anyone become accident prone after seeing Avatar?

Cheers

-Markus

[Eric Lange]

Marc Pesce article does touch on some cogent points even though they are crude or gross over simplifications.

I have worked with stereographic/Photogrammetric and immersive VR data sets and research for over twenty years. For stereo photogrammetric operators, there is substantial testing that is carried out to see if they can “take it” to work many hours a day in a stereo environment. The one thing I have found, is the exact opposite of what Marc Pesce states, is that instead of degrading or ruining people’s stereo acuity, in fact working with “good” stereographic data sets actually improves your ability to handle progressively more challenging ranges of parallax.

The issues and psycho-visual break down occur mainly with incompatible view accommodation versus convergence. I.e. your eye balls swivel in their sockets to converge on an apparent point, but the depth of the point is out of the plane of the view screen that the lenses in the eyeball is used to focusing too. The viewer has to learn to dissociate a particular focusing distance with a particular vergence angle of the eye balls. It’s basically a learnt skill that improves over time.

I think the other thing that causes headaches (which is something that people don’t talk about), is that in nature, when your eyes converge on a point that point is in focus, (for example your thumb at arms length), the back ground will be perceived as a double image, and anything in front of your thumb will also be a double image. The things in the background and foreground are also blurred due to lack of focus (as you are fixing your gaze on your thumb). The stereo visual system is heavily reliant on correspondence matching between left and right images, that have a high frequency content, (i.e. contrasty fine texture and edges). The parts of the perceived image that are out of focus are essentially blurred (i.e. low frequency features), and not rigorously “pattern matched”, never the less the unresolved stereo percept (even as a double image) has a rough depth associated with it. By stark contrast, in the movie theatre, everything is in focus at one plane, and what happens is that your brain is forced to process about twenty times the normal volume of stereo matchable data than it normally would, (as compared to what the system was evolved to do to essentially serve the Paleolithic caveman bumbling around the woods in France, looking for snails and boars to eat). The brain is quite happy to stereo resolve many millions of stereo corresponding points per second, in a sort of stereo spatial gluttony, but after a while, basic channels and bandwidth in the brain DO GET TIRED! [The associated neurons do get wrung out like a sponge of all their dopamine etc.]. By analogy, you can do difficult calculus for a couple of hours, but fifteen hours straight and your brain juices for that capability get totally used up.

I totally agree that sufficient testing has not been carried out to set up various safety guidelines. I am particularly sensitive stereo fatigue myself and yet I have engineers who can work with horrible data sets with their eyeball out on stalks for half the day. However we have a set of technologies that allow one to completely resample stereographic data sources into very comfortable ranges of viewing parallax and still preserve very fine 3d details, so it looks very rich and engaging, (something not possible via conventional techniques). This also allows one to shoot successfully with much wider base separations/IOD. These techniques work especially well for real time environments such as games. This came about as an accidental industrial by product for new computational and spatial paradigms we were researching to vastly improve the computational efficiency of capturing and rendering very complex geometries and real world scenes for stereographic and VR applications. The user (of a 3d TV) can also be given a controller (using our techniques) to in effect dial up or dial down the range of parallax (not just simple HIT). So all of these kinds of problems are cleanly fixable, in a practical way.

Stereo 3d is definitely not going away (now), and if there are class action law suits, bring it on, (our US patent was recently granted), we have technologies to take care of these problems. The question is now, is industry willing to take these types of stereo-physiological issues seriously or will things just bumble on as they are? I personally think the competitive market will decide, and the punters will buy systems and media that have the capability to present themselves in a more stereo comfortable and flexible way. I.e. people like their stereo but they want to control how much stereo effect that they want at a given time, understanding that (like doing math) they can’t push it fifteen hours a day, but at least they are warned that they should take responsibility for how much “parallax” they ingest. It’s a bit like putting a warning on a case of Doughnuts, [Do not eat these doughnuts continuously for fifteen hours straight, every day of your life], all though there will be the occasional “Twinkie” defense in the stereo market place and associated law suits, never the less If extra controls and parallax re-sampling are offered to the user, then it is the user’s responsibility to decide how much parallax they want at any given moment. This is the way forward I believe.

Cheers,
Eric

[Michael Rintoul]

People have been looking at stereo images for over 150 years,.... Are there any cases of stereo blindness as a direct result??? Perhaps the problems arise when you try to move your tv to within one inch of your eye as in the VR headset... try to focus on the dirt on your sunglasses for a couple of hours and then see how you feel.... Not to mention that almost all of the depth sense cues are maintained in 3-D image capture (perspective, occlusion, atmospheric perspective, light gradient, textural gradient, motion parallax, degree of focus...etc) This article is narrowly researched and parallels are draw to 3-D without any merit at all....

[Eric Lange]

There are numerous problems with head mounted displays, but there are optics that allow your eyes to focus to a notional distance for example three feet into the distance. The problem here is that your eyes are forced to focus to that distance, and a dissociation of view accommodation with convergence becomes especially difficult (as compared to 3d tv/monitors). There are many patents out there for HMDs that attempt to change focus in the lenses to actually and physically match view accommodation with convergence. Other techniques try to pattern match some parts of the scene to automatically effect a horizontal image translation to effect a zero parallax condition at the notional point of convergence for the eyes. These systems are fraught with many problems, and are really not that much fun for home use (at least not right now).

As Michael says stereo has been around 150 years, (actually much longer than that for hand drawn examples), but at the same time we must challenge ourselves all the time to create high quality stereo and strive for technical solutions to some of these very real problems. If we ignore these issues, people will get fed up, and stereo will die. But I believe in the relevant industries, that a second wave of products and solutions should take care of things in a fairly substantial and coherent way.

Cheers,

Eric

Foot note:

“Keep doing that you’ll go blind” as the title of the article, is an oblique reference to the Victorian belief that if you masturbate too much you will go blind… and yet stereo was all the rage in the 1880’s,? The Victorians didn’t have much to say about what happens if you use (Victorian) stereographic porn; do you go blind twice as fast? This is a bit tongue in cheek, the article was written by an Aussie I guess with a good sense of humor (from the title of the article). It is an absolute certainty that there will be law suits and ambulance chasers, but it won’t sink an industry and will of course create new opportunities. From my experience working with large VR displays for larger audiences, etc. about one in a hundred and fifty people will have real problems with stereo that lasts for a couple of days, this is a fact that cannot and must not be ignored.

[Dan Hudgins]

There is a serious and major problem with stereo content and screen size.

You CANNOT display the same formated data on the "big screen" and a home TV since on a home TV or computer monitor the distance for points that look at INF would be 65mm and 65mm on a 1000mm wide screen is,

65/1000 or 0.065 the image width.

But if you show that same formated 3D images on a IMAX screen 20,000mm wide the ratio is not the same,

65/20000 or 0.00323

Unlike 2D images that scale to any size screen, the 3D images must be formatted to "fit" the screen size that will be used so that points for subjects at INF have a spacing of 65mm.

If you view formated (65mm) for a 12" monitor (width) then view on a 36" monitor without adjusting the "offset" of the stereo pair you get 65mm * 36/12 =195mm which would make your eyes diverge for INF, something your eyes shoud NEVER do! It gets much worse if you project on a big screen,

65mm* 200000/(12*25.4) = 4265mm

Even though the screen is not as close as the 12" monitor would be and eye spacing (spacing for points at INF) of 4265mm will always make the viewers eyes diverge, and is worse for people who sit close to the screen.

The only solution for using data formated in the same way on different size screens is to set INF at the screen plane so there is 0 disparity for INF and then the eyes can only converge and never diverge.

Otherwise the "stereo" window needs to be smaller than the monitor screen so you can slide the two images around to compensate for the viewing distance and screen size.

For monitoring you should use VR glasses rather than a monitor, or a viewing hood on the monitor rather than glasses.

[Eric Lange]

Dan, using conventional technologies I agree with you.

However there are ways around this, and at least with the technologies we have developed allow one to effectively re-sample the scene real time to new ranges of parallax to suit and optimize for virtually all screen sizes, but 3d rendering capability does need to be in the pipeline, for example a games console or a dedicated (not too substantial) rendering engine for cinematic applications. The sequences can also be re-rendered off line and then stored for play back if necessary. The data sets have to be pre-processed but after that the adjustments can be made real time.

Cheers,

Eric

[Brent J. Craig]

That article makes no sense. The author says that the wikipedia page for 'depth perception' lists 10 different cues that your brain uses to figure out exactly how far away something is, but fails to mention that 3D films use at least 8 of these cues as well!

He author thinks head-mounted immersive VR headsets are the same as watching a movie, and he talks about red/blue anaglyph glasses as if they are commonly used.

Where did he come up with the line "None of the television manufacturers have done any health & safety testing around this."? That's a pretty bold statement to make without any facts to back it up. Did he convince every TV manufacturer to reveal the trade secrets of how or if they test their sets? Bullsh**!

Certainly 3D can be done badly, just like anything else, but done properly by trained professionals it is an amazing experience.

[Dan Hudgins]

Dan, using conventional technologies I agree with you.

However there are ways around this, and at least with the technologies we have developed allow one to effectively re-sample the scene real time to new ranges of parallax to suit and optimize for virtually all screen sizes, but 3d rendering capability does need to be in the pipeline, for example a games console or a dedicated (not too substantial) rendering engine for cinematic applications. The sequences can also be re-rendered off line and then stored for play back if necessary. The data sets have to be pre-processed but after that the adjustments can be made real time.

Cheers,

Eric

I decribed something like that in the 80's but it is not 3D photography once you start relocating the pixels, and you need to resolve the hidden parts.

Also without a standard viewing method, you cannot know in post how the end viewer will see your images, having the viewing device re-format the images is like letting the viewing device re-grade your images, what then is the maker making?

Cardboarding from using compressed imaged in DI and projection makes the 3D value less, because the beauty of 3D in the the fine 3D texture that gets lost with compression that was desigined for 2D images. 3D requires more resolution in each image not less.

I solved the "missing" 3D depth clues issues years ago in some papers I published on how to make "hologram" like video displays with accomidation depth clue and motion parallax in all three axis, that were published by the SPIE in their proceedings, so his blanket statement that there are no ways to resolve the isues is not Comprehensive.