Mathematics of Compressed Sensing Offers Promise of Improved Imaging Systems

May 7, 2007

An emerging application based on compressed sensing signals the possibility of smaller, faster digital cameras—and images of incredibly high resolution. The advance relies on mathematical theory first published in 2004.

Richard Baraniuk and Kevin Kelly, professors of electrical and computer engineering at Rice University, have developed a camera that dispenses with the millions of image sensors of a standard, high-resolution digital camera. Instead, it uses a single sensor to collect just enough image data to allow a novel algorithm to construct a high-resolution image.

The algorithm turns the data into a handful of numbers that it inserts into a giant grid. The computer then fills in the remainder of the grid in a manner that resembles the way people solve Sudoku puzzles. By solving the gird puzzle, a computer reconstructs the complete picture.

Within the next few years, compressed-sensing technology may lead to cell phones that can be used to make poster-size images and MRI systems that are faster than today's scanners by a factor of 10.

Source: Technology Review; Rice University