We deal with vision systems with our company, so this link got passed around this summer. Pretty cool stuff. Here's what my boss said about it:
Colleagues,
Very interesting.
Attached is a technical report concerning the method being used. I suspect the limitation of the technique will be sensor size: each microlens corresponds (roughly, I think) to a pixel, and each pixel must be serviced by a sub-array of sensor elements, so, in order to have decent spatial resolution you will need to have a huge number of sensor elements. This fact, combined with the basic diffraction limit for visible wavelengths, implies large sensors. I also suspect that chromatic aberration will be a challenge.
Then I said "um, yep".
Saw this in (I think) TDDIOT a while back...very cool stuff.
that would be great... I could never figure out that focus thing
Unfortunately, the technology is currently very limited. And per shut's boss - resolution will be a huge issue. This only appears to work, at relatively small sizes. At larger sizes, aberration and overall softness will be an issue (e.g. nothing is really in focus).
It's interesting tech - but currently at least, not going to be very useful for anything other than small web size images. In order to get print sized images, you're going to need at least 3X the regular sensor size - which, for a long time yet, will mean a HUGE cost increase.