DBFNet: A Dual-Branch Fusion Network for Underwater Image Enhancement

Due to the absorption and scattering effects of light propagating through water, underwater images inevitably suffer from severe degradation, such as color casts losses detail. Many existing deep learning-based methods have demonstrated superior performance for image enhancement (UIE). However, accurate correction detail restoration still present considerable challenges UIE. In this work, we develop a dual-branch fusion network, dubbed DBFNet, eliminate degradation images. We first design triple-color channel separation learning branch (TCSLB), which balances distribution by independent features different channels RGB space. Subsequently, wavelet domain (WDLB) discrete transform-based attention residual dense module fully employ information restore clear details. Finally, dual attention-based selective (DASFM) is designed adaptive latent two branches, in both pleasing colors diverse details are integrated. Extensive quantitative qualitative evaluations synthetic real-world datasets demonstrate that proposed DBFNet significantly improves visual quality shows compared methods. Furthermore, ablation experiments effectiveness each component DBFNet.