MRF-UIENet: Underwater Image Enhancement Network Based on Multiscale Receptive Field with Dual Domain Collaboration

Underwater images often suffer from blurring, low contrast, fogging and color deviation due to the scattering and absorption effects of the water medium, which seriously affects their visual effect and application value. To solve these problems, this paper proposes an underwater image enhancement network (MRF-UIENet) based on multi-scale receptive field enhancement. The network adopts an encoder-decoder architecture, captures multi-scale features through the designed D<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>-GateScale module, combines a large-size deep convolutional kernel to increase the receptive field, and enhances the feature expression capability by using the dual-domain (spatial and frequency domains) co-optimization. In addition, a gating mechanism is introduced to control the information flow to further enhance the sensitivity of the model to fine-grained features. The experiments are conducted on the EVUP dataset, and the results show that MRF-UIENet outperforms the existing methods in both PSNR and SSIM metrics, with the highest PSNR up to 26.43 and SSIM up to 0.93, which significantly improves the clarity, contrast, and color fidelity of underwater images.