2/6/2025 | 11:00 AM - 12:20 PM | Regency B

Noise2Image: Noise-enabled static scene recovery for event cameras

Author(s)

Dekel Galor* | UC Berkeley
Ruiming Cao* | UC Berkeley
Amit Kohli | UC Berkeley
Jacob Yates | UC Berkeley
Laura Waller | UC Berkeley

Abstract

Event cameras detect changes of light intensity as a stream of ‘events,’ but cannot measure absolute intensity itself. Hence, they are only used for imaging dynamic scenes. While previous efforts have been focused on filtering out these undesirable noise events to improve signal quality, we find that, in the photon-noise regime, these noise events are correlated with the static scene intensity. We show that by modeling the statistical relationship between noise events and pixel illuminance, we can recover the intensity image of a static scene from the characteristics of noise events. Our method, Noise2Image, can robustly recover intensity images solely from noise events, providing a novel approach for capturing static scenes in event cameras, without additional hardware or specialized pixel circuit modification. Noise2Image enables event cameras to ‘see’ both static and dynamic parts of a scene.