3D Cameras Are About To Go Mainstream

An anonymous reader writes: Vox's Timothy B. Lee reports that everyday imaging is about to take a big step forward as 3D photography finally makes it to prime time. Technological advances in 3D processing algorithms have accelerated at the same time the equipment for taking these shots has become significantly cheaper. Those facts combined mean that we're going to be seeing 3D cameras become much more prevalent very quickly. "If things go according to Intel's plan, within a few years all of our tablets and laptops, and perhaps even our smartphones, will have fancy 3D cameras instead of boring old 2D ones." Throw in the fledgling industries of commercial camera drones and autonomous vehicles, and you have a lot of major companies throwing huge amounts of research money into making cheap 3D cameras work. "The result will be a proliferation of devices, from tablets to self-driving cars, that understand and interact with the world around them."

Comments

Does Anyone Actually Want it?

[tomxor]

Yes it could be done and made cheaply... if it's something that consumers actually want, beyond a gimmicky "My phone has it" selling point.

Maybe i'm just not consumer enough, but i don't really want my photos or video to be 3D, in the same way that film looks better at 24FPS and games look better at >60FPS.

I think high frame rates and depth perception are along the same lines as far as application goes, they bring ultra realism. For things like games, simulations etc that's great. But for many forms of media it seems that lack of realism and it's artistic capacity are somehow entwined, adding ultra realism seems to destroy that. Granted - selfies are tenuously artistic so perhaps this will make it into phones.

Re:Does Anyone Actually Want it?

[Solandri]

Yes you want it. There are two things the humungous DSLR lenses give you: More light so you can capture images in dark situations with less noise. And shallow depth of field.

We''re finally reaching the point I predicted in the early 1990s when the first digital cameras with reduced sensor size came out. That spawned endless debates about what exactly the sensor size did to the depth of field. It turns out when you reduce sensor size, you increase depth of field. This results in photos that look like they were shot with a point and shoot modern digital cameras - everything in the photo is in sharp focus. This happens in the 35mm point and shoot because the lens has a small aperture (ratio of lens diameter to focal length). In digital cameras it happens because they use a tiny fingernail-sized sensor.

To generate creative effects like isolating the subject of a photo from the foreground and background using focus, you need a DSLR with a large lens and large sensor. Would the photo of the Afghan Girl been so striking if the dirty wall of the refugee camp behind her had been in sharp focus?

You can simulate shallow depth of field in software by blurring portions of the photo. But this is usually just a guess based on location in the photo. e.g. Blur the bottom and top third, leave the middle third in focus. It ends up looking rather fake, which is bad unless fake is the effect you're trying to achieve. (That last one's a real scene, it just looks like a miniature because shallow depth of field is also characteristic of photographic miniatures. Your brain has seen it so often that it associates extreme shallow depth of field with miniatures.)

With a sensor which also captures 3D depth info, the sensor and lens size limitation is gone. You can perfectly blur the image in software to simulate any depth of field, from shallow, to deep. Even effects not possible with optical lenses, like non-linear depth of field, are possible. The only remaining reason to lug around huge DSLR lenses is for low-light photography with little noise.

All TV is 3D in your brain

[Anonymous Coward]

The camera isn't the problem, its the viewer.

You capture 3D content, you play it into the brain via a 3D TV and the brain interprets it as 3D scene. When you replay 2D TV images, the brain ALSO interprets it as a 3D scene. When a train drives towards you in a 2D scene, your brain is telling you this is a train coming at you in 3D. Adding some depth changes the path to the brain, but the thing in the brain is still a 3D train.

I noticed when I got my 3D TV, the effect would work for a while and once I got into the movie I would forget I was watching a 3D movie. Switching to 2D did not diminish the movies depth. It did not suddenly feel shallow.

So 3D doesn't add anything.

Only half of what's needed

[Rashdot]

Camera's are only half of what's needed to make them successful. You also need 3D screens without the need for special glasses.

Re:Image quality

[JaredOfEuropa]

I think it's nothing short of amazing how far tablet / phone cameras have come in terms of picture quality, given the crappy lens and tiny sensor. Zoom, low light performance and a halfway decent flash are the obvious things missing from those cameras, but pictures taken in good light conditions actually look very good. Good enough to enlarge and print, and good enough for most people to be unable to tell that the photo was taken with a phone rather than a proper camera.

And nothing beats the camera you have on you.

Re:yay

[Anonymous Coward]

What's wrong with 3D selfies of boobs? Are you gay? Do you hate your country? ARE YOU A COMMUNIST?

Re:Image quality

[javilon]

I think the point here is that with 3D comes the ability to better understand the images. Don't think about producing content. Think about gathering information from the environment. A phone or tablet will always know where it is, even if GPS doesn't work, just by processing the input from its camera (see Google's project Tango). And eventually will understand what the objects it sees are. Think robotics, enhanced reality and many more applications...

For those applications image quality is not relevant.

Re:Image quality

[JanneM]

Think passive near-field 3D-sensors, not holiday snapshots. User position, gestures, navigation, that sort of thing. Kinect-like functions everywhere. Fire phone, but with actual uses.

You could do a lot of subtle UI improvements if you can localize the users in space around the device for instance; you could figure out who is speaking and if they're turned toward the device. No more "Yo, googly Siri-man, what's mein wiener kapiche?"-keywords, as the device can figure out if you're addressing it or not.

HTC EVO 3-D

[crow]

Remember the HTC EVO 3-D? It had a 3-D screen and took 3-D photos and movies. Remember how the revolutionary technology completely took over the market? No, it was pretty much ignored.

I had the HTC EVO 4G which preceded it, and it was a pretty good phone for the day (though Sprint's 4G coverage was horrible--I used it once in the years I had it; too bad they didn't start out with LTE).

3-D has always been a gimmick to attract consumers that has mostly failed. Hollywood is still trying it as a way to get people to have a different experience in the theater from home, but few people seem to care. TV manufacturers jumped on it, but they didn't sell. It's just not something people care about.

Do not want

[AndyMan!]

3D has been "The next big thing" for more than 150 years. Starting with crude Stereoscope viewers in the mid 1800's, a resurgence in the 70's with the first mass market 3D movies, an attempt at another resurgence in the 2000's, and a push by the industry for 3D TV's more recently. Each of these technologies has shown the exact same pattern - a bit of novelty when they're first introduced, then tiredness, and quickly - a clear consumer rejection.

The amount of money tech companies have invested in 3D over the last century is staggering, and the consumer rejection has been consistent. I can't think of a better disconnect between producer and consumer.

So here we are again. We're supposed to get excited over yet another 3D Next Big Thing. No thanks. Just like every example that's come before, I'm perfectly happy and even prefer the current state of my photography. I am an avid first-adopter, but I have absolutely no intention of ever adopting 3D photography.

Re:Do not want

[Ford Prefect]

Wouldn't that make an awesome app, building *real* 3d scenes, and making the models available for export in a variety of formats and with direct-links for popular functions (editing apps, export to popular 3d printing services, etc)?

Look into photogrammetry software like the cloud-based 123D Catch and the defiantly offline Agisoft PhotoScan - they'll turn loads of conventional photos into arbitrary 3D models. The former is probably closest to your request!

I've been playing around with the latter software recently - the required photography is pretty difficult to master, but it's a rather useful tool. Here's a geometry-only render of a statue I scanned as an example - there's a full texture map for the model as well, but this is showing off the frankly implausible levels of geometrical detail you can get from a physical object. (Excuse the noisy crevices - I was shooting hand-held at ~9am in the middle of winter on a cloudy day...) It's terrible at shiny objects (reflections confuse the hell out of it) and system requirements are pretty steep - it'll eat however many CPU and GPU cores you throw at it, and the more memory the better - but the results are well worth it.

Bad summary! No cookie! (3D Computer vision)

[KitFox]

This is definitely a case of picking the worst summary of the source article possible. When I looked at the /. summary, I immediately thought "3D is going out in movies and TV, and haven't we been there with the HTC Evo 3D?". Obviously a lot of other people did too.

So I clicked on TFA. Ahhhhhhhh... Now it makes more sense! From TFA:

We're used to our gadgets being passive objects. They respond to typed or tapped commands, but we don't expect them to be aware of their surroundings.

... As our devices have more and better sensors, they're going to be increasingly aware of the world around them, and will interact with the world and with us in more sophisticated ways.

So other than the really gimmicky "personal drones that can take breathtaking aerial shots", this is primarily talking about computer vision, such as gesture recognition, local environment evaluation, etc.

One interesting side-effect: 3D fakery is harder

[QilessQi]

Photoshopping a single image can be done easily in ways that make the edits virtually undetectable, even for the casual home user. But an amateur attempting to edit two nearly-identical images (e.g., to modify body shapes, or skin tone, or to get rid of unwanted parts of the scene) would almost certainly leave behind inconsistencies that simple image analysis could detect.

...Today, that is.

There will probably be a niche for home-use 3D Photoshop extensions that perform activities like airbrushing, texture duplication, etc. on two images simultaneously in a manner that always results in a clean combination of the two by effectively performing the edits in 3-space. Heck, such technology may already be in use in movie studios that are cranking out 3D movies with live actors that must be composited with generated scenes.

Not 3D cameras

[fyngyrz]

TFA accurately describes using multiple 2D scene acquisitions in order to build a 3D model by trading time and resolution for position.

TFA does not in any way describe "3D cameras."

3D cameras would acquire a 3D representation of subject matter directly. Such cameras do exist; but they are not about to "go mainstream" in any meaningful or accurate sense of the term.

Imagine a 12" cube with numbers on every face. Place a stereo (likely dual sensor / dual lens) camera in front of it, collinear with any one of the six axis. Acquire image. Now, tell me what number is on the face of the cube furthest from the camera.

You can't? Of course you can't. Because you didn't acquire anything even close to 3D data on the object.

Now place the same cube in front of a system that looks at it from, say, 32 directions on a plane parallel to the floor and acquire. Now you can tell me what is on the far side of the cube, because in this case, something somewhat closer to 3D data acquisition was actually performed (and can be used to immediately give you views at angles and distances of much finer granularity than 32.) It's still not actually 3D (what's on the bottom of the cube? The top? For that matter, what's inside it?) but even with the fairly reasonable limits of opacity, at least a system of this kind would be able to present you with the appropriate representation if it was informed that you had moved your viewpoint horizontally or circularly relative to the data's representation on the display device and on essentially the same viewing plane as the camera set.

Re: Not 3D cameras

[Anonymous Coward]

In a consumer video context, "3D" means "stereoptic", not "volumetric". I thought everyone realised that by now.

Re: Not 3D cameras

[fyngyrz]

In a consumer video context, when actual 3D imaging becomes available, there is going to be considerable confusion due to this arbitrary and fundamentally inaccurate usage of the term "3D."

I thought any geek would understand this by now. Stereo is not 3D. It isn't even "2 and 1/2 D" as some like to call it. It's stereo. No more, no less.

more of the same

[fyngyrz]

the article doesn't mention stereo 3D at all

I didn't say it did. TFA (and TFS) title is: " 3D cameras are about to go mainstream. "

3D (X, Y and Z; or width, height, and depth) data is not acquired by these cameras. Therefore, these are not 3D cameras. Therefore 3D cameras are not "about to go mainstream."

It mentions Realsense and Kinect sensors, but it fails to mention how they work.

They work by acquiring stereo 2D data from a fixed viewing angle. That's two acquisitions from almost the same vantage point, which provides a static illusion of depth via capture of parallax. It does not actually contain depth information on a per-frame basis. They are not 3D cameras.

So naturally those depth cameras can't sense the depth of surfaces that are visually obscured, and no, they can't see the back of your cube

Yes, that's exactly what I said. :)

but couple them with accelerometers (for dead reckoning of position & orientation), colour[sic] cameras and machine vision algorithms (for refining that reckoning), and walk around the cube

You can do this with one 2D camera. However, in order to do this, the reconstructed frame rate goes from acceptable to pitiful -- No one is going to be interested in a playback of the Bar Mitzvah that proceeds at one frame per walk-around. A "3D camera" would capture 3D information. Just as an infrared camera captures infrared, a 2D camera captures 2D information, and an ultraviolet camera captures ultraviolet information. It is absurd to characterize a stereo camera pair as a "3D camera", even without considering the bewilderment that will ensue when they actually arrive due to the dimwitted hijacking of the term "3D" by marketing buffoons.

"3D" is short for "three dimensions." That's what it means; that's what it's always meant; that's what it should mean. Suppose I sold you a "3D rendering system" that turned out to only let you specify X and Y co-ordinates for your objects. And when you complained, as you surely would, I tried to feed you a line about how "look, if you simply build, and then render, 2D models of the same object as it would appear from every possible viewing angle, and then display them one at a time, it's almost the same!" your next phone call would probably be to the better business bureau.

chances are you'll actually own (and find a few uses for) such a device yourself in the next few years.

I have owned a Kinect since just about day one. And I am well aware of its nature -- which is not 3D.

The degree of disingenuous wool-pulling over the eyes here is on a level with someone selling you an RGB camera that only captures red and green channels. There's no possible justification for it. None. The resistance to the facts brought about by personal investment in the marketing claptrap is an amazing thing to see -- something that is essentially a particularly rabid form of confirmation bias, where victims of misinformation deny reality because they are unable to admit they've been hoodwinked -- it is one of the things to look for any time propaganda has been drilled into gullible consumers:

o "Hey, bought a new car audio amp, I see!"
x "Yes, it's a Pyle. It's 4000 watts!"
o "No. Dude. It isn't 4000 watts. Someone at Pyle is laughing their head off at you right now."
x "bitch, it says RIGHT HERE that it's 4000 watts!"
o "Sigh."

Essentially the same conversation:

o "Hey, bought a new game motion controller accessory, I see!"
x "Yes, it's a Kinect. It's a 3D motion controller!"

Boobs

[fyngyrz]

Still, selfies of boobs. I think you made a couple of good points. Roundly.