Underwater Image Restoration Based on Marine Snow Removal Techniques

Image capture and analysis in underwater environments have applications in many fields, such as underwater archaeological exploration, deep-sea biological research, oil and gas resource exploration, and path navigation of intelligent underwater robots. Existing underwater image restoration methods usually face problems such as high model complexity and the inability to be deployed on actual underwater operating platforms. This paper proposes an efficient and fast underwater image restoration method called UIRF-Net to address these issues. This method can restore underwater images and remove marine snow phenomena. UIRF-Net makes the deep learning network more lightweight by using modules such as channel extraction prior information and RGB three-channel color enhancement, improving the image color while ensuring the image quality and processing speed. Through experiments on the UIEB-Withsnow dataset, it is verified that UIRF-Net can produce clear underwater images from photos containing complex marine snow noise. This study provides a comprehensive enhancement solution for the field of underwater image restoration, which significantly improves the clarity of underwater photos.