Scientists at Bell Labs have built a prototype camera that uses no lens and a single-pixel sensor. This rather counter-intuitive idea is based around a grid of small apertures that each direct light rays from different parts of the scene to the sensor, and can be opened and closed independently. Using a technique called 'compressive sensing', the sensor makes a series of measurements with different combinations of open apertures, and uses this data to reconstruct the scene in front of the camera. Because there's no lens to focus the resultant image has infinite depth of field, rather like a pinhole camera.

This is the principle of the lensless camera. An aperture array directs light rays from different parts of the subject to a single sensor. This is Bell Labs' prototype camera. The white assembly is the aperture array, and the black box (appropriately) contains the sensor and electronics.
Here's the image of the stack of books shown above, a 302 x 217 pixel file created using 16384 individual measurements. Luckily the lensless single-pixel camera can still be used to take pictures of cats. As long as they're asleep, anyway.

Don't necessarily expect to find this technology coming to a camera near you soon, though. The single pixel design means that a series of measurements has to be made to generate the final image, and the higher the number of measurements, the higher the image quality.

The prototype camera used a 302 x 217 aperture array to produce a 65534 pixel file, which required 8192 or 16384 measurements to make a reasonable quality image (which even then wouldn't stand up to serious pixel-peeping). Scale this up to 12MP - towards the low end of the scale by modern camera standards - and you'd need to make several million measurements to construct an image. As you can imagine, this doesn't make for a quick-shooting action camera - it only works for still life subjects. Meanwhile, zooming would require moving the entire aperture assembly back and forwards relative to the sensor.

Click here for a link to download the original paper (from Cornell University Library)

(via engadget)