Visibility Enhancement for Underwater Robots Based on an Improved Underwater Light Model

Abstract

We propose an underwater image enhancement algorithm for improving underwater robot visibility. Images captured in underwater environments are typically degraded by the effects of absorption, scattering, and noise. Degraded images impede underwater robot task performance (e.g., inspection, detection, and visual simultaneous localization and mapping). In this study, we improve the underwater light model by considering floating particle noise and non-uniform illumination from artificial light sources. Specifically, a systematic underwater enhancement method that includes a floating particle removal algorithm and an image-dehazing algorithm is proposed. Our method is effective for underwater image enhancement applications in real-world scenarios. We compare and evaluate our proposed method with state-of-the-art methods, with an underwater evaluation and a feature-matching performance. The experimental results show that our method yields comparable (and even better) results than state-of-the-art methods.