Underwater image enhancement via multi-angle polarization decomposition

Underwater imaging suffers from severe backscattering in turbid environments. To address this, we propose a robust enhancement method based on multi-angle polarization decomposition. By applying least-squares fitting to dense polarization sequences, we accurately separate the target light from backscattered noise. A simplified de-scattering model and median fusion strategy are further introduced to suppress noise while preserving textures. Experimental results demonstrate that the proposed method effectively eliminates backscatter and significantly improves image clarity and structural fidelity under strong scattering conditions.