Research on muddy water imaging technique based on image enhancement and filtering

In the current underwater survey process of underwater robots, due to the absorption and scattering of light by tiny particles in water, resulting in blurred image edges and reduced texture details, this paper presents a muddy water imaging technique based on image enhancement and filtering for image restoration. Using a fusion image enhancement algorithm based on saturation and saliency for multi-level processing and extraction of different influence features through multiple images, and using a CNN-GAN image filtering algorithm to automate underwater cable image filtering through a two-layer nested neural network with a PSNR value of 29.86 and an SSIM value of 0.9045, The contrast and clarity of the image are significantly enhanced, which is beneficial for the operator's observation of the underwater state.