Polarization total variation method for backscatter suppression and visibility enhancement in turbid underwater environments

In underwater image restoration based on active polarization imaging, accurate estimation of backscattered light is crucial. Factors such as complex underwater scenes or uneven light distribution can exacerbate the spatial non-uniform distribution characteristics of backscattered light polarization, leading to noticeable striping and structural noise in the restored images and thus impairing the restoration quality. To address the non-uniformity of backscattered light polarization, this paper proposes a total variation regularization-based underwater polarization de-scattering method (UDPTV). An energy function is constructed that simultaneously constrains the spatial smoothness of the global backscattered light's degree of polarization and the sparse gradient characteristics of the target structure, and a backscattered illumination non-uniformity metric was adopted to optimize the energy function of total variation regularization. The alternating direction method of multipliers (ADMM) is employed for parameter estimation. In addition, a no-reference image quality assessment metric is introduced as an adaptive strategy to dynamically adjust the regularization weights and step-size parameters, enabling joint estimation of the polarization characteristics of both the target light and backscattered light. Experimental results demonstrate that the proposed method effectively suppresses backscattered light, preserves edge and texture details, and reduces artifacts in the restored image. Furthermore, it exhibits strong robustness in highly turbid and complex underwater environments.