Research on an Improved Shallow-UWnet Based Turbid Underwater Image Enhancement Algorithm

To address the severe color deviation and low contrast issues in images of turbid liquids, a turbid water image enhancement method based on an improved Shallow-UWnet network model is proposed. Firstly, a dataset of such images is constructed by simulating the actual turbid liquid environment. Then, a white balance algorithm is applied to globally correct the color of the initial liquid images. Subsequently, the mapping relationship between the distorted images and normal images is learned using the improved Shallow-UWnet model to enhance the underwater images. Finally, histogram equalization is employed to improve the contrast of the images, resulting in the final enhanced images. Experimental results demonstrate that this method outperforms four other reference methods in terms of both subjective and objective evaluation metrics. It effectively corrects color deviation and enhances contrast and clarity in images captured in different turbid water environments. This method has broad application prospects in underwater image processing, underwater rescue, and other fields involving the processing of images in turbid underwater environments.