When underwater degraded images meet logical stochastic resonance

Abstract

Owing to light attenuation and high background noise, underwater images are significantly degraded, which hiders the development of underwater exploration. However, noise itself can be used to counter noise. In this paper, we apply logical stochastic resonance (LSR) to help detect weak objects from low-quality underwater images. On the basis of analysis of the physical character of underwater images, three models, namely basic dynamical system driven by Gaussian noise, basic dynamical system driven by Ornstein–Uhlenbeck (OU) noise, and dynamical system with extra delay loop, are chosen to study the performance of LSR-based object detection. The main workflow of LSR-based object detection is introduced. To analyze the performance of LSR, we perform explicit experiments and systematically discuss the interplay of additional noise with the system parameters. LSR is proven to be helpful in detecting weak objects from low-quality underwater images. Both OU noise and extra delay loop will help the whole system to maintain stability in a higher noisy background.