A friend is doing renovations and getting his entire ground-floor street-side wall replaced with glass bricks. I told him that with all those bricks giving similar distorted views of the same room, and scattering light in similar ways, there must be some statistical technique to inverse filter the light that comes through to clean the signal and see what's behind the bricks---and that he should watch out or at least consider curtains.

Was I right? Or is this impossible (or NSA level stuff), or can this be done with standard image processing libraries and a little stats?


In theory this is possible, in practice probably very hard, but I wouldn't say impossible. See a recent article by the Silberberg group for example, that demonstrates imaging through 'thin' turbid layers using scattered incoherent light. Also, search for works pioneered by A. P. Mosk on that matter (for example this one and this one). The main challenges are algorithm speed, or, by the time you'll find the correct transmission matrix, it'll change due to thermal fluctuations.

  • 1
    $\begingroup$ Cool, thanks. But thermal fluctuation, really? Isn't the amount of variance in the brick distortion orders of magnitude higher than variation due to temperature? I'm surprised that would matter. $\endgroup$ – Mittenchops Oct 29 '12 at 16:28
  • $\begingroup$ It is the optical wavefront distortions one needs to amend, and these can change significantly for a path difference of some fraction of the optical cycle. So if you illuminate with red light (~600nm optical cycle), think that a fluctuation of a ~100s of nanometers will manifest in some factor of pi phase shift...this won't happen immediately, but if a code takes 10 minutes then it is bound to happen. This is at least true for coherent illumination, see Mosk's PRL papers for more info. $\endgroup$ – bla Oct 29 '12 at 16:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.