A Novel Dataset For Non-Reference Evaluation Of Underwater Image Enhancement

Abstract

Photographs taken underwater are very important for studying marine resources, marine ecology, and environmental monitoring because they give a unique view of the world below the surface. The absorption and scattering of light results in color distortion in underwater photographs. The decline significantly impairs their performance in subsequent applications, such as object detection, monitoring, and identification. Consequently, techniques for enhancing underwater photographs seek to augment visual information and rehabilitate image quality. This work will utilize deep learning, physical models, images, and more methodologies as the foundation for the augmentation tactics presented. Simultaneously, we evaluate relevant metrics for the analysis of datasets and underwater photographs. Our primary objective is to conduct a comprehensive review of the advancements in research aimed at enhancing underwater imagery and its implications for the evolution of superior underwater vision systems.