<p>Underwater image enhancement is fundamental for marine applications yet remains challenging due to complex light-water interactions that degrade image quality through wavelength-dependent absorption and scattering effects. Existing methods often require extensive paired training data and struggle to generalize across diverse underwater conditions. We propose UIEAnything, a novel zero-shot underwater image enhancement framework that integrates automatic white balance preprocessing, physics-guided depth estimation, and an improved restoration algorithm based on underwater light transport theory. Our approach introduces three key innovations: (1) a domain adaptation strategy that bridges the gap between underwater and natural images via physically motivated white balance correction, enabling effective utilization of pre-trained models; (2) an improved Sea-thru algorithm incorporating nonlinear backscatter modeling and adaptive attenuation estimation, accurately capturing the depth-dependent nature of underwater light propagation; and (3) a unified framework that eliminates the need for task-specific training while maintaining physical consistency. Extensive experiments on seven benchmark datasets demonstrate that UIEAnything consistently outperforms state-of-the-art methods, achieving average improvements of 15.3% in PSNR and 12.8% in SSIM. Furthermore, without additional training, our framework demonstrates remarkable generalization capability by successfully addressing other challenging vision tasks involving scattering media, such as image dehazing and sandstorm removal. These results establish UIEAnything as a significant advancement in physics-guided zero-shot learning for image enhancement in complex optical environments.</p>