Scientists have discovered that the human eye has a spooky ability. It can detect “ghost images.”
These are images that are encoded in random patterns, previously thought only detectable by computer. But in a new paper posted online on the preprint server arXiv, scientists in Scotland at Heriot-Watt University in Edinburgh and the University of Glasgow have found that the human eye itself can do the required computations.
“Although the brain can’t individually see them, the eye is somehow detecting all of the patterns, and then keeping the information there and summing everything together,” said study co-author Daniele Faccio, a physics professor at Heriot-Watt University.
Making ghost images
In a regular camera, multiple pixels take in the light from one source, like the sun, to create an image. Ghost images are basically the opposite: They start with multiple light sources in a predictable array with the light being collected by a single-point detector, usually called a “bucket.”
An easy way to envision how this works is to think about lidar, which uses a single-point laser to scan a scene. The detector captures how the light from the laser bounces back from each spot in the scene, which can then be reconstructed into an image.
These ghost images of a four and a six show what the eye sees. Black-and-white patterns are projected against the original images of a four and a six, and the resulting light that bounces back can be processed to reveal the original picture. Researchers once thought only computers could do this processing, but new research shows the human eye is capable of at least some of the process.
But there’s a quicker way to get ghost images, Faccio said. Instead of scanning the scene with a single light source, researchers have found that they can project patterns onto a scene. The light that bounces off the object plus the pattern can then be measured. The difference between that pattern of light and the original pattern projected contains the “ghost image” that a computer can then mathematically wring out of the data. These images look like a faint greyscale representation of the original image.
Computationally speaking, this method of making ghost images involves two mathematical steps, Faccio said. The first is combining the original patterns and the patterns as they appear after being projected on the object. This is mathematically done by multiplying the original pattern against the light signal made by the object and the pattern at each spot. The second is to sum up all those numbers across the whole scene.
“The question we were asking ourselves is, ‘Can the human brain do this?'” Faccio said.
The researchers decided to focus on the second half of the computation, the summing of all the patterns together. To do this, they started by projecting checkerboard-type patterns called Hadamard patterns against the famous photograph of Albert Einstein with his tongue sticking out. They then used a single-pixel detector to collect the resulting light patterns, which they fed into an LED projector.
That LED projector shone the Einstein-plus-Hadamard patterns onto a screen showing the original Hadamard patterns, essentially multiplying the two together. Step one: complete.
The next step was to see what people could see when looking at this summation. The researchers found that when the Einstein-plus-Hadamard patterns were projected slowly, in pulses of 1 second or longer, people just saw black-and-white checkerboards — no ghost images. But as the researchers sped up the projections, Einstein’s goofy visage appeared. The researchers also did the experiments with numerals and letters and found they were legible in the “ghost” versions.
“The blacks and whites will start to disappear,” Faccio said. “They’ll become grayish, and you do actually start to see the image appearing in front of you.”
The reason this works, Faccio said, is that the human eye has a slow refresh rate. It’s not unlike the reason that movies work: When images are flickered on the screen faster than this refresh rate, it creates the illusion of smooth movement.
The eye “is very fast in acquiring the information,” Faccio said. “It’s just very slow in getting rid of it.”
The researchers figured out that the flickering patterns remained in the eye’s “memory” for about 20 milliseconds, slowly fading out over that time. If the 20-millisecond patterns overlap, the eye sums them up like a movie, allowing the ghost image to emerge.
The exciting part of this discovery, Faccio said, is that the ghost-imaging system can be used to study the human visual system. The researchers’ paper is now under review at a peer-reviewed journal. The team’s next step is to find out if the human eye can also conduct the first step of viewing ghost images, perhaps by multiplying together different inputs to the right and left eyes.
Photo Credit: Paul Eddie Yates