interesting idea: shoot now focus later

You just shoot - later at the computer you decide what you want in focus:

http://www.nytimes.com/2011/06/22/te.../22camera.html

This idea has been around since 2005, a company was also formed and tried for several years to commercilize this technology but unfortunayely due to technical limitations it seems very unlikley that it would be used in a normal SLR camera. There were some more interesting ideas developed here between 2006-2008 inclusing a multi-aperture sensor sensor for 3D image capture. The filed of image sensor has matured in the last few years and there is really no more academic research being done. From all of these ideas that existed, only the X3 technology survived which is also struggling now. The rest has just turned into a scaling problem just like microprocessors and memory.

This is the original technical manuscript if you are interested

http://graphics.stanford.edu/papers/...era-150dpi.pdf

Arash,

Thanks for the link, very interesting. The camera that was made for the study needed ~14x14 more pixels, using a 4000 x 4000 pixel array to make 292 x 292 pixel output images. That is a big hit in resolution. Plus matching the f/ratio of the microlens array to the main lens, further limits functionality.

I think most would choose conventional 16 megapixels over a 0.085 megapixel high depth of field camera. But I'm sure there are specialized applications.

Roger

There has been subsequent work on acquiring light fields by other means such as coded aperture and using a mask to build a heterodyned light field camera. Ramesh Raskar and Todor Georgiev are two of many that have done work in this area:

http://web.media.mit.edu/~raskar/Mask/
http://www.tgeorgiev.net/

Though the resolution penalty is much less severe with these techniques, they do require a lot of computational processing (some of which can be done on the GPU) and unfortunately also have a tendency to create nasty bokeh (some of which at least Todor et al. have tried to start addressing).

Its an interesting area of research, but I suspect not ready for commercial applications for a while yet.

But would it be "sharp enough"? :)

And here's the original dissertation.

http://www.lytro.com/renng-thesis.pdf