A multi-branch and multi-scale feature fusion underwater image enhancement algorithm for diverse water environments

In recent years, underwater image enhancement technology has attracted much attention due to its important role in ocean exploration and environmental protection. Compared with the exploration of the marine environment, the study of highland and freshwater-circumstance lakes is poorly understood. Currently, the proposed underwater image datasets are trained and tested on paired datasets from the marine environments, which are overly simplistic and some datasets even contain only the original underwater images. Therefore, it is unknown whether these methods can effectively restore highland and freshwater-circumstance datasets. To address this, we introduce a highland and freshwater-circumstance dataset (P-HFUI) with 2000 pairs of paired images and propose an underwater image enhancement method named Multi-Branch and Multi-scale Feature Fusion Underwater Image Enhancement Algorithm (MM-DUIE) that can restore both marine and highland and freshwater-circumstance. At the same time, we compare the proposed algorithm with the most advanced ones qualitatively and quantitatively. The results show that the proposed model ranks among the top three in terms of Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measurement (SSIM), Mean Squared Error (MSE) and Root Mean Square Error (RMSE), and is more in line with the visual requirements of human eyes.