Underwater images suffer from visibility loss, hue shifts, and patchy brightness due to wavelength-dependent absorption, scattering, and suspended matter. We present an adaptive color-correction framework tailored for turbid water. First, a depth-aware model derived from the extended Beer–Lambert law fuses estimated depth and turbidity to restore edges. Second, a wavelength-adaptive module re-weights RGB channels according to measured attenuation coefficients, retrieving true colours under ideal lighting. Third, a spatial-variance algorithm locally adjusts chroma to eliminate regional casts and equalize tone. These elements are integrated into a unified pixel-level correction function driven by depth, turbidity, wavelength, and local luminance cues. Experiments in mildly to heavily turbid tanks, validated with a full-spectrum colour card, show the method surpasses state-of-the-art algorithms in colour fidelity, detail preservation, and speed. The approach therefore boosts image quality for underwater exploration, ecological monitoring, and cultural-heritage documentation. It also runs in real time on standard hardware without extra calibration.