Single-Pixel Camera

In an ordinary digital camera, the lens focuses light onto a chip containing millions of photodiodes, each of which captures one point of light in every frame. Microprocessors then perform battery-draining computations that compress the data into a smaller form for storage, using an algorithm such as jpeg.

An original image (left) is shown with 85 percent of it removed (center). An image taken with the prototype camera (right) has only 15 percent of the measurements of the original, demonstrating its ability to create high-resolution images comparable to that of conventional imaging.

Richard Baraniuk and Kevin Kelly of Rice University in Houston have developed a technique that con-verts light directly into compressed data. Unlike digital cameras, which process the captured pixels after acquisition, the method, compressive imaging, simultaneously acquires and compresses projections of the image under view without collecting from all the individual pixels.

As presented Oct. 11 at the Optical Society of America’s annual meeting in Rochester, N.Y., they have created a camera prototype comprising a digital micromirror device — a silicon chip covered in thousands of bacterium-size mirrors fabricated by Texas Instruments — and a single photodiode. This system captures random modulations of an image onto one pixel several thousand times in succession.

The mirrors can face in two directions, appearing bright or dark depending on their orientation. Light is collected through the lens onto the digital micromirror device, which then reflects a bright/dark pattern onto the photodiode to make one measurement. The mirror pattern changes hundreds of thousands of times per second, and the photodiode records a new value for each shift in orientation. Newly developed mathematics are then applied to reconstruct a coherent image of the random measurements.

Although it takes several minutes to capture enough data to form a complete image, the prototype takes fewer measurements than there are pixels, which reduces computation and makes the camera more efficient than a conventional digital camera. With only one photodiode, this technology holds promise in terahertz imaging or in applications with wavelengths outside the visible spectrum where current CCD and CMOS devices are blind.