A single-pixel camera

A single-pixel camera has been developed by researchers at Rice University. The device is part of an emerging shift from digital signal processing, in which analog signals are converted into their digital counterparts for processing, to computational signal processing, in which analog signals are fed in some suitable form directly into nonlinear processing algorithms. Instead of data-gathering pixels, the Rice camera uses a digital micromirror device—an array of micromirrors that can each adopt one of two orientations. A lens focuses an image onto the DMD; then the image is reflected by a randomly chosen subset of the mirrors through another lens and focused onto a single photodiode. The photodiode generates a voltage that serves as a coefficient for the particular DMD configuration. The image is sampled repeatedly with different DMD configurations, and the collection of measured voltages is processed to reconstruct the image. Typically, many fewer measurements are needed than the number of mirrors in the array, which leads to savings in data storage and processing. For example, the image of a soccer ball shown here was taken with a 4096-mirror (64 × 64) camera and 1600 measurements. The tradeoff in the new scheme is between data compression and acquisition time rather than between resolution and number of sensors. Because the camera uses only one sensor (a photodiode in the prototype), the researchers say that “compressive sensing” can be adapted for imaging at wavelengths inaccessible to digital photography. The Rice results were reported at the Frontiers in Optics 2006 meeting of the Optical Society of America held in October in Rochester, New York. (See paper FWN3 among the Wednesday abstracts at http://www.osa.org/meetings/annual/program/default.aspx .)