Underwater Image Enhancement and Restoration Method Based on Neural Network

To address the challenges of target detection and recognition in high-turbidity water environments, this paper proposes a neural network-based method for underwater image enhancement and restoration, aimed at improving the accuracy of target detection and recognition under such conditions. We introduce an underwater target image enhancement technique that utilizes multidimensional polarization decoupling and establish a neural network model for restoring polarization images of turbid water, which is based on fully dense connections and pyramid pooling. Experimental verification demonstrates that this method has increased the peak signal-to-noise ratio and structural similarity to 20.08 dB and 0.852, respectively, significantly enhancing the accuracy of underwater target detection and recognition.