Humans can perform ghost imaging, no computer required

Ghost imaging involves capturing an image using photons that have never interacted with the object being imaged. Can you perform ghost imaging using the human eye as one of the detectors? The answer is, apparently, yes. Researchers stared down the barrel of one of their own imaging setups and claimed to observe images. In doing so, they’ve shown that the eye is much like any other camera.

Ghost imaging isn’t all that ghostly

In ghost imaging’s original conception, the quantum properties of light were an essential part of the imaging process. The basic idea was that two entangled photons would be separated. One would head off to bounce off the object being imaged. The other would go to a camera. The photon that bounced off the object would hit a photodiode (essentially a single pixel detector). That click would then be used to trigger the camera to record the other photon, the one that hadn't hit the object. Supposedly thanks to entanglement, an image can slowly be constructed from light that has never been near the object being imaged.

As cool as that sounds, ghost imaging isn't necessarily quantum. Images can be constructed by summing up a sequence of patterns. Each pattern is assigned a value called a weight. If a pattern contributes strongly to creating an image, then it gets a large weight, while patterns that do not contribute get a very small weight. 

As it turns out, the entangled photons were sending a random selection of patterns to the object and the camera. The photodiode clicks more often when a pattern contributes strongly to the reflections from the object. The image at the camera has a large contribution from that pattern because it is triggered by the photodiode.

The role of “quantum” was to ensure that both photons had the same pattern. But you don’t need entangled photons to do that.

Although it still retains the name, ghost imaging is nothing like its original conception now. It can be performed using things like light-emitting diodes rather than expensive entangled photon sources. The patterns are deliberately generated using the same sort of active mirrors that are used in projection systems. About the only bit remaining from the original setup is a single photodiode that is used to detect the reflected light intensity at a single point.  

You don’t even need two beams of light. Instead, the image is reconstructed by a computer. The computer adds all the patterns together, with each pattern being given a weighting (think brightness) determined from the strength of the photodiode signal. You put in light, you detect at a single point, perform some trickery on a computer, and out comes an image.

Taking the computer out of the equation

In this work, researchers have used a clever trick (while clever, in hindsight it seems pretty obvious) to eliminate the computer from the process.

The light is still projected onto an active mirror that generates patterns. The patterned light is sent to the object and detected by a photodiode. The photodiode signal controls the brightness of a second light source. That light source is projected onto a second active mirror that is producing the same patterns.

The patterns from the second mirror are either viewed directly or are projected onto a screen for someone to view. It is then up to the eye to add all the patterns up.

The patterns used here are not a raster scan of the object (though they could be). If you remember cathode ray tubes, then you will know that static images can be impressed upon the brain, even if each pixel is independently lit up in a sequential fashion, rather than all being illuminated at once. You just have to run through them all fast enough that the eye and brain lie to each other and staple all the dots together into one image.