A CNN-Based Framework for Underwater Image Noise Reduction and Visibility Enhancement in Marine Applications

Underwater imaging technology serves critical functions in diverse fields that range from marine biological research to underwater archaeological investigations to oceans exploration endeavors. The underwater imaging process typically produces highly degraded visual output due to multiple factors including light absorption and scattering while also affecting clarity because of particulate matter. We propose a Convolutional Neural Network (CNN)-based technique for underwater image noise reduction which works to enhance image visibility while improving perceptual data. Our system applies deep learning principles to master noise patterns while recovering image clarity. The proposed model trained using underwater images with different levels of noise underwent evaluation on the same dataset. Subjective visual assessment together with objective metrics demonstrated that the CNN delivered noise reduction performance at an accuracy range between 85-90% which defeated conventional noise filtering methods. The developed methodology exhibits substantial progress potential for seabed image quality enhancement tools which subsequently results in better underwater cognitive interpretation processes.