Gradient-guided adaptive polarization fusion for underwater image enhancement in turbid environments

To address the inefficiency and errors caused by manual polarizer rotation in traditional underwater polarization imaging, this paper proposes an active polarization-based enhancement method. It directly generates optimal polarized image pairs via Stokes vector analysis, reducing acquisition time. An adaptive threshold low-pass filtering algorithm is designed to accurately estimate and suppress backscattered light without requiring prior knowledge of the background. Furthermore, a polarization-weighted fusion strategy, guided by a novel Image Quality Evaluation Unit (IQEU) that integrates SSIM, VIF, and contrast weight, is employed to optimize fusion and significantly improve detail clarity and global contrast. Experimental results demonstrate excellent performance across a range of turbidity levels and target types. Compared to existing methods, the proposed approach shows marked advantages in detail recovery, backscatter suppression, and contrast enhancement, highlighting its strong versatility and robustness for underwater detection and high-quality imaging.