Radiometry: Measuring Light as Physics, Not Perception

WHY IT EXISTS To create a standardized, objective way to measure electromagnetic radiation based on its physical properties. Scientific and engineering applications need to quantify the actual energy of light, independent of how a particular observer, like the human eye, might perceive it.

THE MENTAL MODEL Think of radiometry as a physicist's toolkit for light, measuring the actual flow of energy in watts. In contrast, photometry is a psychologist's toolkit, measuring how bright light appears to a typical human, using lumens. Radiometry measures the cause (physical energy); photometry measures the effect (human perception).

HOW IT WORKS Radiometry uses a set of techniques to characterize the distribution of radiation's power in space. It's not one single measurement but a system of related quantities. Key concepts include: radiant flux (total power emitted or received, in watts), irradiance (power per unit area on a surface, in watts per square meter), and radiance (power per area per solid angle, the most fundamental quantity for rendering). These are measured with sensors that have a flat spectral response or a known response curve, allowing for the calculation of pure physical energy, unlike the human eye which has a very specific, non-uniform sensitivity to different colors.

WHEN TO USE IT Use radiometry when the physical effect of light is what matters. This is critical in computer graphics for physically-based rendering, in computer vision for training and operating camera sensors, in remote sensing to analyze satellite imagery, and in astronomy to measure energy from celestial objects. If your system is a machine, it sees the world through radiometry.

WHEN NOT TO USE IT Do not use radiometry when the goal is to describe or specify light for human consumption. For designing architectural lighting, specifying the brightness of a consumer display, or defining color spaces for human viewing, photometry is the correct framework. Telling a user a screen emits X watts/m^2 is far less useful than telling them it has a brightness of Y nits (a photometric unit).

ONE CANONICAL EXAMPLE A self-driving car's camera must interpret the world physically. It needs to know the exact radiant energy hitting its sensor to distinguish a white truck from a bright patch of sky or to avoid being saturated by headlights. The sensor's sensitivity is different from a human's. Using photometry would feed the system misleading data based on human perception, leading to potentially catastrophic failures in object detection.