A New Imaging System Beats the Human Eye 12 Times Over

The human eye is among the favorite talking points of creationists trying to sound science-y. It's just so complex, they say, so perfect in its ingenuity. How could such a device be the product of accidents? They're at least right about one thing: the human eye is a hell of a technology.

Researchers at the University of Grenada have come up with a new system that purports to be 12 times more powerful than either the human eye or conventional camera techniques. Their technique, which is described in this week's edition of Applied Optics, involves a new light-sensing technology that's capable of extracting the full color information of a scene (or the pixel of a scene) without the need for filters.

Filters are a crucial piece of the typical architecture behind most cameras, whether it's a smart-phone camera, super-high end DSLR, or anything else interested in the visible spectrum of light. First, there's some black and white sensor, and stacked above it, there will be layers of color filters corresponding to the usual red-green-blue (RGB) primary colors.

Per single pixel of the image, only a single color can be sensed. Obviously, a single pixel can deliver more information then that, and this is where algorithms come in. While some pixel might have registered just red, software will step in and sort (interpolate) it all out using information gathered from other pixels via the process of demosaicing.

So, a digital camera might register something like what's above, but that's not the image seen by the photographer. In between the final rendered image and the single-colored tiles/pixels of a raw image file is essentially a math problem.

The Grenada researchers, led by optics professor Miguel Ángel Martínez Domingo, skip this system entirely, replacing conventional RGB color filters with a new type of sensor known as a transverse field detector (TFD). First proposed in a 2008 paper by scientists at the Polytechnic University of Milan, TFDs replace filters with a certain variety of material that allows photons (light particles) to penetrate different distances into the sensor depending on their energy.

Imagine blasting darts at a dartboard; if you really whip it, the thing is going to be harder to pull out because it penetrated deeper into the cork. It's sort of like that.

As different energies (wavelengths) of light correspond to different colors, the TFD can determine what color a given photon represents based on how deeply it sinks into the sensor material. The crucial fact is that this sort of sensor can register all colors at once.

The existing Foveon X3 digital sensor uses the same general principle, albeit in a cruder form. The Foveon still relies on the usual three-color RGB scheme, with an array of three different sensors at three different depths. TFDs can wrangle 36 different channels of color information. "We can modulate the depth at which the photons in each colour channel are collected," said Martinez Domingo in a statement provided by the University of Grenada. "This offers the possibility of fine tuning the way in which these sensors turn the light they receive into electric signals."

The human eye itself is limited to the three-color scheme, which is where the "12 times better" hook comes from. So what difference does it make? What good is an image that's too good for us to even appreciate? The Grenada researchers suggest schemes ranging from the sensing units in driverless cars, medical imaging systems, and spotting counterfeit bills. The human eye is great, but, after all, it remains hardly perfect.