Performance metrics for neuromorphic imaging

Event cameras offer significant advantages over conventional frame-based sensors, including high temporal resolution, wide dynamic range, and efficient power consumption, making them valuable for dynamic imaging tasks across many applications. The absence of event camera benchmarks and performance metrics, however, hinder their use in high-end imaging systems - specifically for complex systems that require expensive development cycles and high performance. This work discusses the importance of performance metrics to evaluate event camera capabilities at pixel and sensor levels, independent of algorithmic processing. A key contribution is the introduction of a time constant for target detection, τ_detection, derived from changes in event firing rates between a target and its background. This metric provides an intuitive and computationally simple measure of system performance, applicable across a range of tasks. A modular simulation pipeline is constructed to test this metric under varying optical, scene, and sensor conditions, including effects of motion, optical blur, and line-of-sight vibrations. Results demonstrate the relationship between detection performance, scene parameters, and imaging system characteristics. The findings underscore the potential of event cameras in time-critical detection tasks and pave the way for developing unifying spatial-temporal metrics that address feature separation and target tracking.