<p>Underwater images are often degraded by color cast, low contrast, and noise resulting from wavelength-dependent light absorption and scattering. These degradations are unbeneficial for analysis and applications. Moreover, the inherent image noise tends to be amplified during enhancement, producing a noticeable granular texture. To address these challenges, an underwater image enhancement method via color-noise decoupling and reconstruction (CNDR) is proposed. Specifically, the underwater image is first decomposed into color components and detail-noise components in the frequency domain. The color components are subsequently enhanced through a dual-stage attenuation compensation (DAC), which effectively corrects color cast while accounting for turbid and low-light conditions. For contrast improvement, CNDR employs a dual reconstruction strategy. The reference-guided contrast reconstruction enhances global contrast using color components, while the feature-preserving contrast reconstruction improves local contrast by leveraging detail-noise components. Comprehensive experiments conducted on three benchmark underwater datasets demonstrate that CNDR achieves superior performance compared with existing state-of-the-art techniques, particularly under turbid and low-light conditions. The proposed method also shows strong potential for engineering applications, such as the segmentation of tidal stream turbines for fault diagnosis.</p>