Ghost imaging is a curious phenomenon that has had numerous physicists scratching their heads in recent years.
It works like this: take two beams of entangled photons and aim the first at an object. The transmitted photons from the object are then collected by a single pixel detector.
The second beam is aimed at a CCD array without ever having hit the object.
It turns out it is possible to reconstruct an image of the object–a so-called ghost image–by matching the data from the two detectors, even though the single pixel detector has no spatial resolution.
When this was first demonstrated in 1995, everybody was amazed by the strange power of quantum entanglement.
But later, various groups showed that entangled beams weren’t necessary at all and that ordinary light from a pseudothermal source would do the job just as well.
While interesting, that doesn’t actually rule out the possibility that the two beams may be correlated in some entangled-like quantum way, however.
So the question of whether quantum entanglement is responsible or not has remained open. Until now.
Yaron Silberberg and pals from the Weizmann Institute of Science in Israel, have carried out an ingenious experiment that settles the matter.
They use only one beam, which they use to illuminate the object, and collect the transmitted photons using a single pixel detector. They then calculate theoretically what the second beam should look like and combine the single pixel data with this “virtual beam”.
And get this: they still see a ghost image. That’s a 2D image from a single pixel detector! And a pretty convincing demonstration that quantum entanglement cannot be responsible.
The question now is: what kind of classical information processing allows the reconstruction of a 2d image from a single pixel sensor? That’s a real puzzle.
Ref: arxiv.org/abs/0812.2633: Ghost Imaging with a Single Detector
This is very clever, but it doesn’t seem all that spooky to me. The computer is sending random patterns to the SLM and correlating the detector results. It’s sort of a holographic version of this demo: http://graphics.stanford.edu/papers/dual_photography/
and from the abstract it sounds like pretty much a direct optical recreation of this: http://www.osti.gov/energycitations/product.biblio.jsp?osti_id=758929
and there’s at least a conceptual parallel with “compressed sensing”: http://www.dsp.ece.rice.edu/cs/
Wim,
I don’t believe it is just a conceptual parallel to compressed sensing. It is compressed sensing. They recover the original signal with 16000 measurements out of possible (300×300=) 90000 possible pixels. They do that by using some incoherent (random) mask. Except for the reconstruction technique which I am not familiar with, it is CS. I’ll mention it on my compressed sensing blog (http://nuit-blanche.blogspot.com).
Thanks for using the keyword, I would have never found it otherwise.
Igor.
This is extremely interesting! see the reconstruction movie of the image at the group’s website:
http://www.weizmann.ac.il/home/feori/Misc.html
What constrains the image to 2-D? Does some mechanism exist by which one could trade say, frequency or spatial resolution for depth of field?
Actually, the image reconstruction in Computational ghost-imaging is 3-dimensional, as we show in the out-of-focus reconstructed image.
In principle one can reconstruct a couple of transparency/images placed at different depths, as long as the each transparency does not interfere the beam propagation too much (e.g. by using fluorescent samples which absorb only a small fraction of the light.