3D Imaging (Score 1)

The news article motivating this thread arises from this article in yesterday's issue of Science: "Visible Cone-Beam Tomography With a Lensless Interferometric Camera," by D. L. Marks, R. A. Stack, D. J. Brady, D. C. Munson Jr., and R. B. Brady, Science Jun 25 1999: 2164-2166 There is a better news article in describing the work in Science itself: "3D Camera Has No Lens, Great Depth of Field," by Daniel Radov, Science Jun 25 1999: 2066-2067. These are available through Science's web site, but a subscription is required. Science offers a 1 day subscription to the web site.

The paper uses a combination of interferometric imaging algorithms (which image with infinite depth of field ) and computer tomography algorithms ( which combine infinite depth of field images to produce 3D models) to produce a 3D image of a plastic toy. Opacity is not a problem due to the linearity of the imaging process.

As several posters have noted at this site, pinhole cameras also have infinite depth of field. We wrote an paper about using pinhole cameras for tomography in Optics Letters last year. Unfortunately, the depth of field of a pinhole camera comes at the expense of resolution. This is not true of interferometric cameras.

"Interferometric" refers to measurements of cross-correlation functions to isolate intensity contributions from different points in the object space. The algorithms used are very similar to those used in radio astronomy.

"Tomography" means slice (tomo) plotting. "Computer" is the C in CAT scan or CT. Current usage applies tomography to most 3D imaging schemes. "Coherence tomography" is a point by point scanning scheme which, ironically, is not tomographic at all. Tomography allows parallel data acquisition, which ultimately leads to real-time 3D video and holodecks. Real-time 3D was not demonstrated in the Science article, however, because it requires a dense sensor array. See http://www.phs.uiuc.edu/Beowulf for progress on this front.

Concern about DARPA and big brother issues is unnecessary. A sensor array has no better chance of seeing inside opaque objects than a single camera. Anyway, why should big brother waste a lot of effort to get information people will volunteer in exchange for supermarket dicount cards.