基于红蓝通道先验主动光源偏振水下图像复原算法

Underwater image restoration algorithm based on red and blue channel prior active light source polarization

摘要:

针对光在水中传播距离短、损耗大造成图像信息损失严重、对比度低、色彩畸变问题，提出一种基于红蓝通道先验的主动光源偏振水下图像复原算法（ALPUIR-RBCP），首先，研究光线在水体传播时的偏振特性，建立水下光学偏振模型；在此基础上，研究不同通道光信息差异性的先验知识和偏振度之间的映射关系，构建多通道光学偏振成像模型；进一步，结合后向散射光偏振度的光敏感性以及光分量间相关性，计算光照分布图和互信息，实现自适应偏振度估计，提高偏振度计算精度，增大目标信息光比重，从而提高图像复原质量。实验结果表明，对比传统的偏振光算法以及传统图像增强和复原算法，本算法不仅在主观上具有良好的水下图像复原效果，同时在各项评价指标中均有提高。

Abstract:

An underwater image restoration algorithm using active light source polarization based on red and blue channel prior is proposed. Aiming at the problems of serious image information loss, low contrast and color distortion caused by short propagation distance and large loss of light in water, firstly, the polarization characteristics of light traveling in water are studied, and an underwater optical polarization model is established; on this basis, different channels are studied. The prior knowledge of the difference of light information and the mapping relationship between the degree of polarization, a multi-channel optical polarization imaging model was constructed; further, combined with the light sensitivity of the degree of polarization of the backscattered light and the correlation between the light components, the illumination distribution map and Mutual information, realize self-adaptive polarization degree estimation, improve polarization degree calculation accuracy, increase the proportion of target information light, thereby improving image restoration quality. The experimental results show that, compared with the traditional polarized light algorithm and the traditional image enhancement and restoration algorithm, this algorithm not only has a good underwater image restoration effect subjectively, but also improves in various evaluation indicators.