A micro camera the size of a grain of salt can take clear, full-color shots equivalent to those generated by ordinary cameras more than 500,000 times larger.

The secret is disguised in over a million “eyes” the size of an HIV virus that cover the camera’s metasurface.

The micro camera has the potential to revolutionise in-body and brain imaging, autonomous driving, robotics, and other sectors where computer vision is crucial.

How micro cameras work?

Micro cameras that use metasurfaces already exist, but their image quality is limited by physics.

A standard camera employs a series of lenses to focus light and capture sharp images. However, the camera must be large enough to accommodate those lenses.

One method for shrinking a camera is to use “meta-optics,” which employ exceedingly small nanostructures capable of capturing and re-emitting light. According to Nerdist, each of these small structures syphons up the photons that produce light, and the metasurface then converts those photons into a signal that a computer can understand. Imagine them miniature antennas.

Micro cameras had previously been developed using this meta-optics approach, but the results have been disappointing, according to Newsweek – images seem like coloured streaks viewed through cataracts, and the cameras can have a small field of view.

“It’s tough because there are millions of these minute microstructures, and it’s not clear how to build them in an ideal way,” Princeton PhD student and study co-author Ethan Tseng explained.

“We could make individual surfaces into ultra-high resolution cameras, so you wouldn’t need three cameras on the back of your phone anymore.”

– Felix Heide

To solve the problems, the Princeton micro camera team turned to science’s Swiss army knife: machine learning. Shane Colburn, a researcher at the University of Washington and study co-leader, developed an AI model that could evaluate the configurations of the tiny antennas on the micro camera metasurface. Because there are so many antennas and their relationships to the light they capture are complicated, the simulation requires “huge quantities of memory and time,” according to Colburn in the release.

The researchers may produce images that are significantly better than the present cutting edge by combining the physical qualities of their micro camera with this mathematical optimization.

While the margins are slightly blurred, a flower still looks like a flower; according to the researchers, the field of view of the micro camera and the high-quality photographs are new high water marks.

“Although the approach to optical design is not novel, this is the first system that utilizes surface optical technology in the front end and neural-based processing in the back end,” said Joseph Mait, former chief scientist at the United States Army Research Laboratory, in a press statement. Mait was not a member in the study.

The big picture

The researchers believe their micro camera may be dusted on any surface and turn it into a camera. (As Nerdist points out, this isn’t the most comfortable concept.)

“We could make individual surfaces into ultra-high resolution cameras, so you wouldn’t need three cameras on the back of your phone anymore,” Princeton computer science professor and research senior author Felix Heide stated in a press release.

More positively, the micro cameras could assist improve current endoscopic surgical methods — in which small equipment, including a camera, are used to do surgeries as non-invasively as possible — and open up new paths for brain imaging.

“In the future, we can think of radically alternative approaches to manufacture things,” Heide added.

Info source – Freethink