I sat down last night to get my head around how the light-field-camera system works, because it looked conceptually difficult ...I couldn't work it out at first run through, I was bewildered.
It took a few different approaches and looking at a Lytro video of their first series of interesting cameras, for the penny to drop. You can think of the technique as on the one side being like the compound eye of a fly, fusing together many individual shots in the 'brain' to make a map of the immediate world ahead of you. Or you can think of it as grabbing more information from the light rays which enter the camera, which in traditional photography are simply trained onto a flat 2D image with much of what the lens can capture being simply reduced to a simple, easy to interpret image.
Think then this. Today we are very used to 3D CAD images, while I work often with 2 D output for actual practical engineering and manufacturing purposes. Plan view, side view and so on.... It is easy on the computer to flick between these two , the three dee for conceiving how it will look, operate and fit during installation and then the 2D for the nitty gritty of measurements, angles, threads, welds and so on, where 3 D would just be a mess of info overload. Usually the 3 D is made from the 2 D traditional drafting, and then the CPU brain makes the 3 D view.
This is a similar principle for light field cameras, where a traditional lens is followed up by an array of micro lenses, the fly's eye , which then allow for more information to be captured in terms of all the focal planes available ie many slices of depth of field, and then also the multiple stero/optic capture where a slight 3 D image can be recomposed. Each micro lens splits the detail up a little more from all the rays coming such that the rays of light render more information or more traces if you like onto the sensor. It has been done in film too, but is not practical for viewing before the advent of serious computing power.
We got used to the stunning quick time style composite images which were created by multiple camera capture, allowing for unbelievable 3 D images with about a 40 degree walk around. The rolling stones video was the first I can remember, and I wondered, how did they do that!??? Basically on a single sensor you can achieve a slight 3 d stereoscopic effect due to the distance across the chip being able to capture very slightly different angles of view
Depth of field contol is then achieved in post processing with Lytro's software suite for example and that is the biggest benefit for general photographers. For macro photographers, the 3D view also helps them but really it seems to be a bit gimmicky for general photography, it is only a few degrees.
The compromise of this composite single capture image is the eventual resolution though. The first lytro camera has a resolution as any one still jpeg image, or just 1.2mpx . This is fine for a VGA monitor but makes for a limited image size on anything bigger, or a poor quality large image. Whereas in fact the information processing is far, far higher than that of a normal 2 D camera due to all the extra computations, which links a camera with a reasonable ouuytput of 8mpx into needing serious on board processing in order to capture even one image per second and represnent it as a flat jpeg on its read monitor screen.
Howwever, Lytro have of course found out that it is not consumers who really gain from this small 3D effect in still images, it is the space imaging and military gun sight government research that this system can find most value and thus earn Lytro the most cash. For me I can see two other major areas if it can be made to work with microscopes and endoscopes, for bio/medical imaging and for inspection work in engineering.
I dont think we will all be using Lytro cameras any time soon, but perhaps we will see a mobile manufacturer licensing in the lens array because then VGA quality in maybe a 2mpx image is worth the bother for the small screen size to overcome the normal DOF limitation and offer a quirky three dee image.