Underwater Image Enhancement Method Based on Vision Mamba

To address issues like haze, blurring, and color distortion in underwater images, this paper proposes an underwater image enhancement model called U-Vision Mamba, based on the Vision Mamba framework. The model first uses a U-shaped network encoder to extract multi-scale features. To aggregate these features, we introduce a novel multi-scale sparse attention fusion module into the network. The decoder then processes and refines the enhanced features to produce high-quality underwater images. Experimental results show that our algorithm effectively addresses image blurring and corrects color distortion. It performs well in both subjective and objective evaluations, demonstrating the model's effectiveness and robustness.