CENet: Color Selection and Edge Enhancement Network for Underwater Image Enhancement

Underwater image enhancement technology plays an indispensable role in marine scientific research, underwater engineering, and underwater security. However, the underwater image enhancement is frequently met with some obstacles, primarily owing to the changeable underwater environment. Building upon these insights, this study introduces an innovative model-Color Selection and Edge Enhancement Network (CENet), which significantly improves the performance of underwater image enhancement through multiple modules. Specifically, it includes the Dense Nested Channel Selection Module (DNCS), the Edge Refinement Module (ERM), and the Gradient-Aware Corrector (GAC). DNCS enhances the model's attention to more severely attenuated color channels while fusing multi-scale features through a dense nested structure. ERM fuses the shallow-level detail information and the edge information of the predicted image to enhance the image details. GAC refines the edge texture features of the image through the gradient map. CENet demonstrates competitive results on the LSUI dataset, with SSIM reaching 88.13 and PSNR reaching 24.21.