U-Shape Swin-Convs Transformer for Underwater Image Enhancement

Due to the scattering and attenuation of light into the water, the underwater image usually appears with color distortion, blurred details, and low contrast. To address these issues, this paper proposes a novel end-to-end model based on convolutional neural networks and attention mechanisms for underwater image enhancement. Specifically, to overcome the limited long-distance dependency capability of the U-shaped structure based on pure convolutional networks, an attention mechanism is embedded into the U-shaped structure to improve the network's ability to capture global dependencies. To reinforce the network's attention to the color channels and space areas with more serious attenuation, a new U-shaped structure is constructed through integrating with a Hybrid Channel Transformer with Multi-scale Feature Fusion module specially designed for underwater image enhancement task. Extensive experiments using the Underwater Image Enhancement Benchmark, Large Scale Underwater Image Dataset and Stereo Quantitative Underwater Image Dataset demonstrate the effectiveness of our method.