CSAIL Event Calendar: Previous Series

Definition of "computational imaging":

(1) Replacing complex, precision physical apparatus such as lenses,
mirrors, magnets, collimators, Anger cameras etc. with computing.
Computing is rapidly becoming cheaper and more powerful --- physical
apparatus is not.

(2) Computing images from sensor information when physical apparatus
cannot be used to make images, as for example, at wavelengths where no
material can usefully refract or reflect radiation.

Examples of computational imaging:

(*) Synthetic Aperture Microscopy (work with Dennis Freeman, Michael
Mermelstein, Jekwan Ryu, Stan Huang etc.). Focus on the resolution in
the illumination system rather resolution in the imaging system, as in
the traditional approaches. Achieve subpixel resolution. Switch to purely
reflective optics for extreme short wavelengths.

(*) Coded Aperture Imaging (work with Richard Lanza, Alberto Accorsi
etc.) Image with gamma rays, neutrons, X-rays --- where lenses and
mirrors cannot be used. Replace rats with mice. Image cargo and
contraband at a distance.

(*) Exact Cone Beam Reconstruction (with Xiaochun Yang and Biovisum,
Inc.) Area sensors promise better use of X-ray tube output and hence
much increased scanning speed, as well as isotropic spatial
resolution. But, once again, the old algorithms do not apply, and
yield only approximate kludgy solutions. Finding optimal "orbits".

(*) Diaphanography --- "diffuse optical tomography" (with Xiachun Yang).
Recover absorption distribution in highly scattering media from
measurements made on the surface of a volume. How to "invert
the heat equation" when radiation transport is dominated by diffusion.

For more information, please visit
http://www.ai.mit.edu/people/bkph/courses/cs294-6/compimagdetail.html