Underwater image enhancement using multiple processing techniques based on DCP, CLAHE, CNNs, and U-Net

In the last decade, interest in the underwater world has increased due to the abundance of resources and abundant species of aquatic organisms and their reliance on them as a source of food or energy. It was necessary to prepare the necessary conditions to make what is underwater visible naturally, which is difficult to achieve due to the loss of color in the blue and red channels, in addition to darkness, fog, refraction, and dispersion. All of these things require us to do our best to make what is underwater easy to control and monitor. For this reason, work was done to develop a fusion algorithm for many techniques, starting with removing fog, improving luminance, reducing noise and preserving edges, then obtaining fine details, then multi-level analysis to enhance lighting, then building the trained model to extract image features and improve them for vision, highlighting final details and improving sharpness, then performing the accurate evaluation process using quality measurement standards between the original and final images, Which led to obtaining good results for the proposed method compared to modern algorithms in terms of results with the standard quality criteria used (PSNR, SSIM, RMSE, VIF).