Enhancing Underwater Images with KAN and Attention Mechanism

Capturing clear underwater images is significantly hindered by the complex behavior of light in aquatic environments. Scattering, absorption, and attenuation of light degrade underwater visibility, leading to issues such as color distortion and diminished contrast in images. In this study, drawing inspiration from KAN and attention mechanisms, we introduce a novel, reliable, and lightweight Underwater Image Enhancement (UIE) network, termed KRANet. This network leverages the synergy of spatial and channel attention mechanisms, along with residual connections integrated with KAN, to improve image quality. Experimental results demonstrate that KRANet outperforms nine other UIE methods, exhibiting superior objective performance and consistent visual enhancement across various real underwater images.