Underwater Image Restoration Based on Color Polarization Imaging and Deep Learning

Polarization imaging technology demonstrates significant potential for underwater applications. This work presents a novel approach that integrates deep learning (DL) with polarization imaging to restore degraded underwater images. Experiments were conducted using a division of focal plane (DoFP) polarization camera to capture color images, followed by the reorganization of polarization information from RGB (red, green, blue) channels for neural network training. A hybrid loss function is proposed to optimize the restoration process. Experimental results indicate that the proposed method achieves superior image recovery performance compared to existing approaches utilizing single loss functions.