State-of-the-Art Deep Learning Methods for Underwater Image Enhancement

Enhancing underwater images is a crucial task for numerous purposes such as marine exploration, underwater robotics, and aquatic ecology. Underwater images are degraded due to the complex propagation and scattering of light in water, leading to low contrast, color distortion, blurring, and noise. Compared to conventional methods, deep learning offers superior performance and flexibility for underwater image enhancements in recent years. This research presents a comprehensive examination of deep learning models aimed at enhancing underwater images, including exploration of methods for enhancing underwater images, network architectures, metrics for evaluation and benchmark datasets. We also provide a comprehensive analysis of the cutting-edge techniques on several datasets, emphasizing their respective strengths and limitations. Furthermore, we discuss the systematic investigations for underwater image enhancement, open issues with challenges, and suggest some promising directions for future research.