What do sudoku and digital imaging have in common?

What do Sudoku and digital imaging have in common?

By Sandy Cosser – 22nd February 2008

If you know an engineer you probably know a perfectionist. Engineers are never satisfied with “good enough” or “satisfactory”. As a rule they aren’t even satisfied with “the best possible” and continually come up with new ways to improve on projects or products. It’s an attitude they share with scientists and is in no way a bad thing, as they drive technological developments and keep human kind on its toes. Richard Baraniuk and Kevin Kelly, in true engineering style, have come up with a method that will revolutionise digital imaging, improving the resolution of medical scanners and cameras.

In current technology, each image sensor in a digital camera designates a number to the bit of light that strikes it. When all of the numbers are pieced together, they form the image of the picture taken. An onboard computer has to compress the picture and in the process gets rid of most of the numbers. It’s a complicated and laborious process that uses a large chunk of battery power.

Baraniuk and Kelly’s technique eliminates the need for images to be compressed. Their cameras use only one image sensor to collect the barest minimum of information. A specially designed algorithm then uses the information to reconstruct a high-resolution image. This compressive sensing technique needs only a fraction of the information required by existing cameras to create complete images.

The algorithm used is designed to turn the visual data it receives into only a few numbers that are randomly placed on a large grid. The sudoku analogy now comes into play, as the algorithm goes about completing the grid in much the same way that we solve sudoku problems. When the grid is complete, the photographic image has been successfully recreated.

Compressive sensing will vastly improve the functioning of MRI systems, as they will be able to capture images ten times more quickly than current scanners. Kelly believes that the improvements in MRI applications will be practically viable in about two years’ time. Consumers will have to wait another five to ten years for the technology to become available, in the form of minute mobile phone cameras that provide high resolution, poster-sized images. Thus proving that size does indeed not matter.

Recommended site:

http://www.technologyreview.com/read_article.aspx?ch=specialsections&sc=emerging&id=18293