ShutterEvents: Shutter Modulation for Static-Dynamic Scene Recovery

Event cameras capture changes in log intensity asynchronously, enabling high temporal resolution and dynamic range for dynamic scenes. However, because they respond only to intensity changes, they are fundamentally insensitive to static or slowly varying content, leading existing event-to-video methods to rely on camera motion or hallucinate missing structure. Prior efforts to recover static information either exploit Poisson noise events—requiring long integration times—or employ mechanical shutters to convert event cameras into intensity sensors, sacrificing temporal resolution.

We introduce ShutterEvents, a lightweight sensing framework that enables simultaneous recovery of static and dynamic scene content using a low-cost ($7) programmable LCD shutter. By imposing controlled temporal modulation on incoming light, static intensities are transformed into predictable brightness variations that generate informative events even under a stationary camera. We show that the induced mapping between event statistics and scene intensity is injective, enabling a learning-free reconstruction of static structure directly from event measurements.

To separate shutter-induced responses from true scene dynamics, we propose an event decomposition algorithm that leverages the global temporal synchronization introduced by the modulation signal. Experiments on both simulated and real data demonstrate that ShutterEvents recovers meaningful static structure while preserving dynamic fidelity, without long integration times and with minimal hardware overhead.