Optical performance benchmarking turns the vague idea of a “good camera” into measurements that laboratories can repeat and compare. Under controlled illumination, we test how little light a device can detect, how finely it can resolve detail, how much noise it adds, and how wide a range of brightness it can handle. These numbers are the common language behind every technology programme at the Photonics International R&D Center. They allow engineers to separate genuine sensor progress from an attractive image or a single headline specification and they help partners understand whether a system will work in the real world [1].
For intensified night vision, no single measurement tells the whole story. Researchers look at figure of merit (FOM) limiting resolution multiplied by signal-to-noise ratio alongside photocathode radiant sensitivity, halo diameter, auto-gating response and modulation transfer function (MTF). These laboratory values describe how the tube behaves; range models then translate them into the practical question of how far away a person or object can be detected, recognised or identified. The classic Johnson criteria remain widely used, while TOD and TTP methods provide more modern scene-based predictions [2], [3]. Digital cameras are commonly characterised with EMVA 1288, which provides a shared method for quantum efficiency, temporal noise and dynamic range [1]. The field is now moving beyond isolated datasheet numbers towards full measured distributions and performance linked to realistic scenarios.
At the Photonics International R&D center, we place analogue filmless tubes, EB-CMOS hybrids and AI-enhanced digital imaging chains on the same optical ‘starting line’. Each device sees the same calibrated low-light sources and target sets, and its results are analysed with the same statistical treatment of uncertainty. This matters because processed digital imagery can look cleaner even when the underlying sensor has not collected more useful photons. We also measure performance by wavelength rather than relying only on photopic, or daylight-weighted, values. Two regions are especially important to our roadmap: solar-blind UV-C at 200–280 nm and the near-infrared band around 1060 nm. Protocols, environmental records and traceable calibration make every result repeatable by another laboratory.
The hardest comparisons are still ahead. Intensified analogue imagery and digitally processed output do not fail in the same way, and AI enhancement can hide noise, invent detail or change contrast. UV-C imagers still lack a consolidated benchmark standard, while many established tests use static charts even though real scenes contain movement, changing light and atmospheric effects. Our next step is therefore to turn laboratory practice into a shared method: contribute to European measurement working groups, publish protocols that partners can reproduce, and develop tests for dynamic scenes. The goal is practical European measurement infrastructure for low-light and dual-use photonics so that a result measured in one laboratory means the same thing in another.
REFERENCES
[1] European Machine Vision Association. EMVA Standard 1288 Characterization of Image Sensors and Cameras (Release 4). Available: https://www.emva.org/standards-technology/emva-1288/
[2] Exosens/Photonis. Figure-of-merit and DRI performance documentation for 4G-class image intensifiers (2023–2024). Available: https://www.exosens.com/products/image-intensifier-tube-4g
[3] NATO STANAG and military DRI modelling literature: Johnson criteria; TOD/TTP range-prediction methods.
[4] Photonics21. Light Driving the Future—Photonics Strategic Research and Innovation Agenda 2028–2034 (2026). Available: https://www.photonics21.org/