Paper Title
Enhancing Underwater Imagery - Advanced Image Quality Improvement Using Cyclegan for Marine Research and Photography

Abstract
UIE(Underwater image enhancement) has been a challenging task because of its unique distortions and degradations present in underwater(UW) environments, including low contrast, colour distortions, and blur. Traditional image enhancement techniques often fail to effectively restore fine details and improve image quality under such conditions. In present years, DL(deep learning)techniques, mainly Generative Adversarial Networks(GANs), shown promising results for addressing these challenges. This paper presents and evaluates several advanced GAN-based models for UIE, including RAUNE-Net, WaterNet, UGAN, and FUnIE-GAN. These models leverage the power of adversarial learning and image-to-image translation to enhance UW images without the requirement for paired datasets. We conduct extensive experiments on several UW datasets, including UIEB100, EUVP_Test515, and Ocean Ex, to estimate performance of each model in terms of Peak Signal-to-Noise Ratio(PSNR) as well as Structural Similarity Index(SSIM). Results demonstrate that RAUNE-Net outperforms other models in both objective metrics and qualitative visual assessments, making it the most robust solution for UIE. The study also discusses the potential of incorporating attention mechanisms and residual learning for further improvement in image restoration tasks. Keywords - Generative Adversarial Networks(GANs), Underwater Image Enhancement; CycleGAN; Image Restoration; Attention Mechanisms; Deep Learning; Peak Signal-to-Noise Ratio; Residual Learning; Structural Similarity Index; Unsupervised Learning.