Stanford researchers cram 12,616 tiny lenses into a 3D camera
By Nilay Patel 
posted Mar 19th 2008 7:07PM
posted Mar 19th 2008 7:07PM
With the megapixel race already past the point of noticeable benefit to consumers, it looks like the next camera arms race will be the number of lenses your rig sports -- a team at Stanford is working on a 3D camera that uses 12,616 micro-lenses to generate high quality 3 megapixel images with self-contained "depth maps" that measure the distance to every object in the frame. The system works by focusing each lens above four different overlapping sensor arrays, which work in concert to determine depth -- just like your eyes. Unlike similar systems, the Stanford rig is able to use that data to create a depth map without lasers, prisms, or even complex calibration, which will allow the team to shrink the tech down to compact and cellphone camera size. Once it's ubiquitous, the teams says depth map information can be used to do anything from enhancing facial recognition systems to improving robot vision, but there's still a long way to go -- the team has just started trying to work out how to manufacture the system.
Wow, this is really quite neat! Can it model NURBS for me? Huh? Please?
All I want to know is "Will it blend?"
Well, I know what "Will it play Doom?" means but I also keep hearing "Will it blend?" can someone explains that to me?
Check out www.willitblend.com (not sure if I can put websites here but here we go!)
FIRST
I for one welcome our doom playing, blending overlords.
Happy now?
Just imagine ONE DECADE from now - it will probably be a Video Camera
OK- I read this and immediately thought of the seen in Minority report where he's playing the 3d-ish movies of his kid at the beach on the hologram thing. If it could shoot at a quick enough framerate- this would be able to shoot video just like that- w/ real depth, but from a single POV. Now, to design a reasonable system for playback...
i remember when 3mp with one lens was amazing. now they can do it with 12,616 simultaneously
Focusing this with shaky hands may prove a bit of a problem. Imagine the waiting time needed to focus all those lens.
That's what I was thinking haha!
Combine this with a 3 axis inertial sensor and you could generate a point cloud model of a room just by panning the camera around! Cool implications for high-speed 3D modeling.
With technology like this, you don't even need an inertial sensor to generate a point cloud. With the addition of a 3d depth map, it's very easy to identify points to track the 3d movement of the camera.
Does anyone have a link showing the output of one of these?
Frank, this sounds like it is based around plenoptic cameras (or integral photography), which have been around for awhile. If you visit http://graphics.stanford.edu/papers/lfcamera/, they have a good video that shows some of the uses of plenoptics circa 2004-2005, such as the "3D" capture, rotating viewpoints, and digital refocusing.
So, if plenoptic cameras have been around for so long, why is this one making the news? According to the press release, it sounds like this is the first time that the imaging, sensor elements, and electronics have been integrated together onto a dedicated chip -- as opposed to modifying a current camera to do the plenoptic thing.
You might also check out coded aperture photography as another interesting way of encoding 3D information into a single photograph (http://groups.csail.mit.edu/graphics/CodedAperture/, also with cool video), or digital holography as a more quantitative research tool.
What an epic day, first we see that Verizon is going to offer an any device service, and now we've got depth mapping cameras. Bring on the robotics revolution! The next 10 years are going to make the .com boom look quaint.
Here we go again....starting from....3mp
The Google maps party van should get some of these.
"Once it's ubiquitous, the teams says depth map information can be used to do anything enhance facial recognition systems to improving robot vision..."
Where do I start on the grammatical problems? This is why the internet is ruining the english language - when even journalists won't proofread their work.
Bloggers are considered Journalists these days? Damn this pesky internet and all of it's confusing tubes!
ah, i believe that English should be capitalized. as should ...Ass Hat!
How ironic.
and without the.
The more pixels are shrunk, the less light each pixel will have to process at a given shutter speed.
The data on surrounding pixels could be aggregated, but the noise per pixel will be very high. Since pixels being digital by nature, low levels of data would be lost unless these pixels are as light sensitive per area compared to how small they are.
This camera will be very hard to use in low-light, or any movement.
Looks like "The Fly" eyes are here.
Anyway, how are we going to see the 3d? On a special screen, through glasses or what?
And here is the paper.
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4114959
[warning, link requires subscription :) ]
I'd love bring a video camera based on this tech into work and film the whole place. Then I'd turn it into a Doom map. Fraggin co-workers (and my boss) FTW!