Enhancement of Underwater Images and Restoration using Deep Learning

Learning-based systems have intermittently demonstrated potential in enhancing underwater photographs. These advancements enable us to efficiently study, visualize, and comprehend the universe. The maintenance of ecosystem equilibrium relies on underwater monitoring; however, due to the refraction and absorption of water, underwater images often exhibit diminished contrast and color in captured images. To improve underwater image quality, image enhancement strategies are necessary. This proposed framework tackles the issues of underwater photography. The proposed method includes Red, Green, and Blue Histogram Shifting (RHGS), Dark Channel Prior (DCP) scale Image Processing (MIP), Underwater Local Adaptation Processing (ULAP), and Contrast Limited Adaptive Histogram Equalization (CLAHE) models based on Rayleigh scattering are some of the techniques used. We employ CLAHE for pixel correction. We utilized the DCP method for removing hazing, the MIP algorithm for optimal object visibility, and the ULAP algorithm for light scattering and backscattering. The effectiveness of the suggested method is assessed by comparing its performance against existing state-of-the-art methodologies, specifically in terms of PSNR and MSE.