Researchers in the US have unveiled a 1 gigapixel camera, which has about five times as many pixels as today’s best professional digital cameras and nearly 100 times as many as a compact consumer camera. Moreover, the camera has a much smaller aperture than other gigapixel devices – meaning that, unlike other sensors, this latest camera pushes the fundamental resolution limit of optical devices. The team has also shown how the device could be used in several applications including surveillance, astronomy and environmental monitoring.
In the past, gigapixel images have be formed by stitching together 1000 or more megapixel images, or by scanning a sensor across a large-format image. Acquiring “snapshot” gigapixel images is trickier, but a few options exist or are in development. One of these is the 3.2 gigapixel digital camera that will sit within the Large Synoptic Survey Telescope (LSST), an optical device currently under construction in northern Chile.
In theory, the smallest details resolvable by a lens are limited by diffraction, and the larger the lens – that is, the greater the aperture – the smaller the details it is possible to identify. A 1 mm aperture should be able to resolve about 1 megapixel, while a 1 cm aperture should be able to resolve 100 megapixels. The LSST, which will have an aperture of several metres, should be able to resolve not just images of gigapixels, but of terapixels.
In practice, however, large lenses struggle to reach their diffraction limit. One problem is that big lenses are more likely to suffer from aberrations, which smudge the focusing, particularly around the extremities. Lens designers get round this problem by introducing more lens elements and reducing the field of view. Nevertheless, these devices still generally fail to get close to the diffraction limit.
Pushing the diffraction limit
Now, David Brady of Duke University in North Carolina and colleagues claim to have created a high-resolution camera that approaches the diffraction limit. Known as AWARE-2, their camera has an aperture of just 1.6 cm yet offers a resolution of 1 gigapixel. For visible light, that is half way to the diffraction limit of 2 gigapixels.
AWARE-2 uses a “multiscale” design, where one spherical objective lens projects a coarse image onto a sphere. On this sphere, an array of 98 microcameras, each with a 14-megapixel sensor, refocuses and samples the image. “The design approach is directly analogous to the development of supercomputers using arrays of microprocessors,” says Brady. “We build supercameras using arrays of microcameras.”
Drop in price
The Brady group’s approach works because it replaces one big camera with a composite of tiny cameras, which are less prone to aberrations. It also means a cut in cost: the mobile-phone market has brought the cost of sensors down to about $1 per megapixel, which suggests cameras of the AWARE-2 design could one day cost just $1000 per gigapixel. Brady thinks it should be possible to manufacture cameras for less than $100,000 per gigapixel by 2013. “We hope that our systems will reach $10,000 per camera within 5 to 10 years,” he says.
However, not everyone is impressed. Engineer David Pollock of the University of Alabama in Huntsville believes the AWARE-2 camera suffers from “a lack of focal length”. Shorter focal lengths tend to have lower “f-numbers” – that is, wider relative apertures – resulting in less thermal noise. On the other hand, shorter focal lengths mean less magnification: the subjects recorded on the camera sensor appear very distant.
Minimum focal length
This is true of the AWARE-2 camera, which offers a very broad field of view of 120°. But Brady does not consider this feature a disadvantage. “In my experience, lens designers universally regard the ability to design to the minimum focal length possible as a positive thing,” he says.
“I would say that this debate gets to the heart of the innovator’s dilemma of the battle between good and excellent,” he adds. “Current high-pixel-count imagers for aerial photography and astronomy are very good systems and their designers are naturally hesitant to believe that a better approach is possible…It’s a fun battle to fight.”
The team has also shown how the camera could be used in a number of different applications. In one example, the camera acquired an image of a lake in North Carolina that was then analysed to determine how many swans were on the lake. In another example, details such as car licence-plate numbers and individual faces were picked out of a surveillance image.
The work is described in Nature.