An Information-Theoretic Perspective on the Challenges and Advances in the Race towards 12μm Pixel Pitch Megapixel Uncooled Infrared Imaging

a-Si (amorphous Silicon) microbolometer FPAs (Focal Place Arrays) with TEC-less (without Thermo-Electric Cooler) and shutterless capabilities have become the technology of choice for low cost, high resolution and low SWaP (Size, Weight and Power) uncooled LWIR (Long Wave Infrared) cameras used in mobile applications. Over the past 10 years, a-Si microbolometric FPAs have seen a steady reduction in pixel pitch from 45 mu m to 17 mu m as well as an increase in pixel count from 160x120 to 1024x768. Next-generation arrays are projected to feature 12 mu m pixel pitch and resolution up to 1440x1080. However, microbolometer technology scaling has detrimental effects on pixel performance and the imaging system's optical complexity, which does not always yield a better infrared image quality. In this paper, we describe, from an information-theoretic perspective, the benefits of using computational imaging technologies and more specifically pupil function engineering to compensate for the optical resolution and noise sensitivity problems caused by shrinking pixel geometry in microbolometer FPAs. Computational imaging is a developing field in which the image acquisition process is shared between the optics and post-capture digital processing (cf. encoding-decoding scheme).