Low-light imaging - Fundamentals and perspectives

The field of low-light technologies is defined by its applications. Historically, this field has been dominated by military applications. Since the late 1950s, the needs of the armed forces to carry out their missions at night have been the driving force behind these industrial developments. The fundamental aim is to provide mankind with the best possible night vision capability, with the aim of coming as close as possible to his day vision capability. This motivation can, of course, be applied to a number of civilian applications. In addition to the field of night vision, there are many other scientific applications that make use of these technologies. To name just two - and it would be unrealistic to be exhaustive - we can mention two extreme cases: astronomy and microscopy.

Low-light night vision is defined by a spectral range in which photon fluxes are low or very low. Initially, because of human capabilities, this domain was limited to the eye's spectral band. Then, naturally, the limitation of photon flux and technological capabilities led to the expansion of this spectral band to explore and exploit the limits provided by the terrestrial environment. We won't go into the thermal infrared field here, which, day and night, requires the manipulation of large photon flows, while being only partially sensitive to solar illumination. The thermal infrared domain enables us to produce images dominated by temperature contrasts. Thermal images of scenes are therefore very different from images perceived by human vision, which is sensitive to variations in the reflectivity (or albedos) of objects illuminated by natural or artificial sources. We're going to focus here on this area, where scene images are dominated by albedo contrasts, both day and night. This limits the spectral range at long wavelengths to 3 μm. At short wavelengths, the range is limited to the near UV, at around 0.3 μm, by atmospheric absorption.

This article is divided into two main parts. The first deals with the analysis of the various contributors to night illumination, scene contrast and the fundamentals of the signal-to-noise ratio in low-light imaging. The second part reviews technological alternatives for detection and imaging in this spectral range. Particular attention is paid to CMOS arrays. An analysis of characteristics, particularly sensitivities, concludes this article. Finally, we conclude with the best current choices and future prospects.