Researchers at the University of Washington and Princeton have created a full-color, high-resolution camera that can capture images as small as a grain of salt. This is the next generation of metasurface tech.

Although tiny cameras have been known for their ability to see inside the body, and aid scientists and doctors in various ailments, past implementations of compact cameras based on metasurface technology have been limited to producing fuzzy images and limited field of view.

Researchers state that sensors with submicron pixels are possible, but further miniaturization is prohibited by fundamental limitations of conventional optics. “Traditional imaging systems are composed of a series of refractive elements that correct aberrations. These bulky lenses place a lower limit of camera footprint. Another fundamental obstacle is the difficulty in reducing focal length. This causes greater chromatic aberrations.

These researchers seem to have solved these problems. The team published a paper on Nature that was summarized by Princeton University. It shows that the camera can produce “crisp” full color photos that are comparable to a conventional compound camera, which is 500,000 times more powerful. The new camera is called a “neural-optic” system by the team.

The metasurface is dotted with approximately 1.6 million posts, each about the size of an HIV-infected virus. Princeton explained that each post is unique in its geometry and acts as an optical antenna. Combining these unique cylinders with machine learning algorithms, which interpret light hitting each one and combine the data to create high-quality images, they are called machine learning algorithms.

Below is an illustration from the paper showing what the neural nanooptic camera can capture in comparison to other methods.

The researchers explained that they made comparisons with a hyperbolic metaoptic for 511nm and a state-of the-art cubic metaoptic from Colburn. “Additional experimental comparisons against other single-optic or meta-optic designs can be found in Supplementary Note 11. Ground truth images were obtained using a six-element composite optic, which is 550,000x more volume than the metaoptics. Our complete computational reconstruction pipeline is available at real-time speeds and takes only 58 milliseconds to process a 720px x720px RGB capture.

Researchers are currently working on adding computational capabilities to the camera. In addition to improving image quality, they also want to add object detection and sensing capability that could be used in medicine and robotics.

“We could transform individual surfaces into cameras with ultra-high resolution so that you don’t have three cameras on your back anymore but the entire back of your smartphone would become one big camera,” Felix Heide (the study’s senior author, and an assistant professor in computer science at Princeton), says. We can imagine completely new ways of building devices in the future.