Etc-Net: a space-adaptive swin transformer-based method for underwater image enhancement integrating edge sharpening and color correction

Underwater optical images are important means of conveying information about underwater scenes, contributing significantly to underwater environmental monitoring, terrain surveying, and biological detection. However, the quality of underwater images is often compromised by the absorption and scattering effects of water on light, resulting in challenges such as low contrast, blurriness, and color shifts. Aiming at resolving these challenges, the present study introduces ETC-Net, an underwater image enhancement network utilizing an encoding-decoding architecture. The proposed approach includes an input-side edge information fusion module to enhance initial image informativeness, a spatially adaptive Swin Transformer module serving as a fundamental unit in both the encoder and decoder to mitigate blurriness and an output-side per-channel color correction module to address color shift phenomena. Experimental results on multiple datasets demonstrated the exceptional performance of ETC-Net in both reference and no-reference evaluation metrics, effectively improving the visual perception of underwater images. The code is published at https://github.com/lqjw81/ETC-Net.