Seafloor photographs generally experience severe degradation due to optical properties of light, like absorption and scattering, leading to color anomalies, reduction of contrast, and loss of details. Such vision loss conditions create severe issues for marine science and computer vision-based ocean exploration missions. Thus, an underwater image restoration method is proposed using a wavelet-based U-Net architecture. By incorporating wavelet decomposition into a deep learning framework, the proposed model can extract multiscale spatial information efficiently, preserving the image clarity with structural and textural details. Wavelet-based model is trained on the Large-Scale Underwater Image (LSUI) and Enhancing Underwater Visual Perception (EUVP) datasets, both of which offer diverse underwater scenes captured under various lighting conditions and water types. The model’s performance is evaluated quantitatively in terms of qualitative analysis metrics. For the LSUI dataset, the proposed method achieved a peak signal-to-noise ratio (PSNR) of 23.47 dB and a structural similarity index (SSIM) of 0.9127. Evaluation on the EUVP dataset yielded a PSNR of 23.59 dB and an SSIM of 0.8572.