Russell Brinkworth

Bug eyes.
New video software based on a fly's visual system can take a normally dark image (bottom) and enhance an important detail, such as a person obscured by a plant (middle), and then highlight that detail (top).

Spying With a Fly's Eyes

For more than 150 years, photographers have wrestled with the problem of exposure. Attempting to take a picture containing a wide range of light intensity meant sacrificing part of the image--either washing it out or plunging it into deep shadow--and thus losing detail. The problem is particularly acute in security applications, where the inability to differentiate human faces hidden in shadows can be disastrous. Now researchers think they may have found a way to overcome this challenge and perhaps create a new generation of video cameras that can see clearly no matter what the light conditions.

The team took its inspiration from the humble housefly. Although a fly's eyes contain multiple lenses, its 1-milligram brain assembles all the signals into a single coherent image that is remarkably detailed, although more softly focused than what the human eye can see. When the researchers inserted microelectrodes into living fly brains, they recorded the electrical impulses from each neuron that connects to the eyes. The resulting video representations of what the insects were seeing revealed details even in the brightest and darkest areas. "When it comes to seeing, even a tiny insect brain can outperform any current artificial system," says team leader Russell Brinkworth, a physiologist at the University of Adelaide, Australia.

Based on the findings, Brinkworth's team developed software that works like a fly's brain. "The solution was not, as most people think, to put more pixels in the camera," he says, "it was to make the pixels smarter." The software quickly establishes light values independently for every pixel, either enhancing or diminishing the signal to preserve detail, then compresses the data, as a fly's brain does. The program runs on a computer chip called a high-speed analog very large scale integration device, or aVLSI, which can process such information very quickly but requires little power. Placing the chip between the camera lens and the image sensor, the team has produced a prototype system for the U.S. Air Force, which is interested in the technology for airborne surveillance.

The findings just go to show that studying insect visual systems can produce "clever and unexpected ideas" on how to improve digital imaging technology," says Mandyam Srinivasan, the director of the Centre for Visual Sciences at the Australian National University in Canberra. "This research will pave the way for the design of novel, biologically inspired approaches for the guidance of miniature, autonomous aerial vehicles with a variety of industrial, surveillance, and security applications."

Related site

Posted in Math