Defogging imaging using second-order correlations in the time domain

Abstract

The traditional space-domain McCartney model simplifies fog as a time-invariant medium, as the fluctuation of light field introduced by time-varying fog is a noise for optical imaging. Here, an opposite finding to traditional idea is reported, i.e., the noise introduced by time-varying fog can be eliminated by itself. The space-time McCartney model is proposed to study the second-order correlations of the time-varying scattering light field in the time domain. We theoretically and experimentally demonstrate that the noise photons, which cause image degradation, lead to the absence of stable second-order correlations, while the signal photons, which produce ideal images, are opposite. The noise photons and signal photons are distinguished by measuring the temporal second-order correlations when fog is time-varying and time interval is longer than the coherence time of the light field, thereby reconstructing high-quality defogging images. Distinguishable images can be directly obtained even when the target is indistinguishable by conventional cameras, providing a prerequisite for subsequent high-level computer vision tasks. The space-time McCartney model provides a theoretical framework for studying the light field properties of time-varying media, and offers promise for anti-interference imaging.