Space-Time Nonlocal Metasurfaces for Event-Based Image Processing

TL;DR

This paper introduces a novel passive metasurface capable of performing mixed spatio-temporal differentiation for event-based edge detection, enabling ultrafast, low-power image processing suitable for neuromorphic cameras.

Contribution

The work presents a metasurface design that performs advanced spatio-temporal differentiation, advancing passive optical computation for dynamic image analysis.

Findings

Successfully demonstrated event-based edge detection using the metasurface.

The metasurface can be tailored to detect objects moving at specific speeds.

The design is compatible with standard fabrication techniques.

Abstract

Analog computation with passive optical components can enhance processing speeds and reduce power consumption, recently attracting renewed interest thanks to the opportunities enabled by metasurfaces. Basic image processing tasks, such as spatial differentiation, have been recently demonstrated based on engineered nonlocalities in metasurfaces, but next-generation computational schemes require more advanced capabilities. Here, we tailor nonlocalities in space and time to design a metasurface that can perform mixed spatio-temporal differentiation of an input image, realizing event-based edge detection with a passive ultrathin silicon-based structured film compatible with standard fabrication techniques. The metasurface detects the object edges only when the object moves, and its design can be tailored to selectively enhance objects moving at desired speeds. Our results point towards fully-passive processing of spatio-temporal signals, for highly compact neuromorphic cameras.