Courtesy of the Make Blog, here is Mike Golembewski’s page on Scanner Cameras. You can get more details on building a similar camera from Jason Yang’s thesis, A Light Field Camera for Image Based Rendering or Wang and Heidrich’s Design of an Inexpensive Very High Resolution Scan Camera System. Good stuff.
I’m a bit frustrated by this, because the RCA webserver seems to be swamped with traffic. Apologies if I’m ranting. I’m guessing that the content is something like Andrew Davidhazy’s material on strip photography, although I’d be amazed if it was as extensive or well presented. http://www.rit.edu/~andpph/
To tell you the truth, I think Heidrich et al sound a bit too pleased with themselves, given that they haven’t really done anything imaginative. If they had user a scanner to make a super-high quality lensless pinhole/slit camera, or a very cheap light field camera, that would have been cool. But just sticking a flatbed scanner on the back of a conventional camera? Must try harder.
Incidentally, I wonder if you could use a lenticular sheet to turn a scanner into a lightfield camera. Someone (below) made a light-field viewer along those lines, but I haven’t heard of a camera that works that way.
http://graphics.stanford.edu/~billyc/research/autostereo/initial.html
Editor’s note: Thanks for the great links Theo. I had seen Davidhazy’s stuff before, but not the other. I had seen the hex array stuff used before in a Siggraph paper, if you consult Fresnel Technologies price list, you find that they cost about $55 a sheet (#300, 8 inches by 10 inches, with .1 inch lenses). At 300dpi, you only have resolution of about 30 pixels accross, or about 700 pixels in area. That’s comparable to most computer icons, not exactly high resolution. I believe the current generation LIDE scanner can acquire 1200dpi, which would be about 120 pixels accross. Still not amazing resolution, but potentially interesting.