Mastering the Challenges of Compressive Sensing

Compressive Sensing cameras are not built like typical cameras. Typical cameras use a Focal Plane Array (FPA) with a detector diode per pixel. The FPA in a typical camera can sense, all at once, all of the pixels of an image.  But there is a downside to this FPA approach; FPAs get expensive for non-visible light wavelengths, such as Shortwave Infrared (SWIR), where exotic materials such as InGaAs are needed to fabricate the FPA.

Compressive Sensing makes an interesting tradeoff, using measurements over a short period of time and computational imaging processing to dramatically reduce the amount of expensive detector material required.

Building a robust Compressive Sensing camera has required InView to invent new technology. We have faced a number of challenges. For example, with a Compressive Sensing architecture we make a series of optical measurements over a short period of time, rather than all once as is done with a FPA, to learn about the incoming image. We use that time-series of measurements to computationally construct the output image. But what happens if the background light level is varying as we make that series of measurements?  Those changes in the background levels get mixed with the measured image information and can significantly degrade the quality of the output image.

Recognizing that challenge, InView filed patent applications in early 2010 describing the fundamental light measurement problem and the foundational inventions needed to deal with the problem. We recently received notice from the patent office that our essential technology will soon be covered by two patents.

The images below, taken with an InView210 SWIR camera, show the value of our basic IP.

text w before-after

Applicable US patents include:  8,570,405  and  8,570,406.

Tagged with: , , ,