Underwater image enhancement presents unique challenges due to blurriness, color distortion, and reduced visibility caused by water properties. In this paper, we propose WRSU-Net, a W-Net architecture combined with a Residual connection and Small U-Net Block (ReSU-Block) to address these issues. The W-Net’s encoder-decoder design, enhanced by the ReSU-Block’s ability to capture multi-level features, significantly improves the clarity and structural accuracy of underwater images. By utilizing a combination of Smooth L1 Loss [3] and SSIM Loss, our approach achieves superior results in comparison to existing methods. Experimental results demonstrate that our model outperforms other state-of-the-art techniques on the LSUI dataset, achieving a PSNR of 28.624 and SSIM [13] of 0.925.