Miniature foveated imaging camera could lead to high-resolution swallowable pill cameras or better drone vision.
Do you want a camera inspired by the way that eagles and other winged predators see the world, with super-high-resolution at the center and then lower resolution around the edges? Of course you do.
Fortunately, that’s exactly what researchers at Germany’s University of Stuttgart have developed with a new miniature camera, capable of being 3D-printed onto a chip to form a multi-aperture camera.
This “foveated” imaging system (named after the fovea area of the eye, which offers the greatest sharpness of vision) can be used for tasks in which only the central section needs to be incredibly detailed. The setup itself consists of four different lenses, each with a different focal length and a field of view ranging from 20-70 degrees. It works by capturing images at low and high resolution at the same time, and then creating a foveated image when they are overlaid.
Foveated images can be useful for a variety of different reasons. Two big ones are compression and the speed of image retrieval or analysis. This could make it a useful tool for future autonomous cars, in which snap judgments need to be made based on image recognition.
“What is new with this work is our ability to take advantage of the strengths of a new optics fabrication process,” lead author Simon Thiele told Digital Trends. “This approach allowed us an extensive variation of parameters on a very small scale, and enabled us to 3D print four different complex lens systems with different focal lengths in one step directly onto the image sensor.”
Due to this high level of miniaturization, Thiele said the work has a number of intriguing possible use-cases.
“The most exciting applications in my opinion are in the fields of endoscopy and drone vision,” he continued. “If we are able to supply swallowable pill cameras with optics similar to those of eagles or humans, we could improve the imaging inside the human body. Similarly, a small drone with eagle vision would probably outperform one with a normal camera.”
As to what’s next for the project, Thiele said that team must further improve the optics and print them onto smaller imaging sensors. “At the moment, the camera size is limited by the available lateral chip sizes,” he concluded. “To be useful for consumers, our approach would have to be ready for mass manufacturing. This is not the case at the moment, but possible in principle with enough funding.”