Enhancing Underwater Images: Automatic Colorization using Deep Learning and Image Enhancement Techniques

Color restoration is challenging in underwater imagery. In underwater environments, red light is easily absorbed by water, resulting in the disappearance of red light at a few meters depth. However, the rapid advancement of deep learning techniques has provided possibilities for underwater color restoration. In this study, we developed an underwater color restoration system based on two deep-learning network architectures. We enhanced the quality of underwater images and explored the use of generative adversarial networks (GANs) for automatic coloring. We compensated for the absorbed red light in underwater environments to improve the color representation of restored images. We combined the WaterNet and neural colorization networks and retrained them to achieve the goal of color restoration. We evaluated the results by observing color palettes and conducted corresponding analyses using k-means, CIE-2000, and percentage differences in color channels.