Digital adaptive optics for event-based cameras

Digital adaptive optics (DAO) is a self-referencing homodyne interferometry technique for mitigating atmospheric turbulence in passive imaging systems. Research so far has focused on reconstructing images from conventional, framebased CMOS imaging sensors. The work presented here will combine event-based cameras with homodyne interferometry techniques, creating a new sensing methodology. Event-based cameras are a new class of sensor whose pixels operate asynchronously and transmit signals only when a relative intensity change above a certain threshold has been reached, producing a stream of timestamped events encoding pixel location and polarity. While event cameras have lower resolution than conventional frame-based cameras, their pixels have higher dynamic range and lower latency, creating an advantage in persistent surveillance applications. Pairing an event-based camera with a DAO system poses certain challenges as the existing calibration and reconstruction process operates on single image frames. The baseline approach explored in this paper is to aggregate events occurring over a given time interval and generate a synthetic frame representation. While this reduces some of the benefits of event cameras, namely their low latency pixels, it enables rapid exploration of the technology with existing toolsets. In this work, we explore this tradespace to find a balance. This presentation will report on experiments undertaken with two event cameras: the DAVIS346 from iniVation (346 x 260), and the Prophesee EVK4-HD (1280 x 720). The DAVIS outputs both grayscale intensity frames and asynchronous events. The frames can be used directly during DAO calibration, and as a comparison point for the synthesized event frames.