Underwater optical imaging plays an irreplaceable role in fields such as marine exploration, resource development, environmental monitoring, and underwater archaeology. However in the underwater environment, light is affected by scattering and absorption, which not only leads to blurred image details and reduced contrast, but also severely limits the imaging range, and further makes it difficult for traditional technologies to acquire scene depth information. Since underwater scattered light inherently exhibits partial polarization characteristics, polarization technology can separate the target light and reflected light by analyzing the polarization state of light, thereby effectively suppressing the scattering effect of turbid water. Nevertheless, existing polarization underwater imaging methods mostly focus on "enhancing image clarity" and neglect the hidden depth cues in polarization images. Although binocular imaging is suitable for underwater 3D perception due to its simple hardware and low energy consumption, in turbid water, objects with low texture are prone to failure in feature point matching, resulting in wrong depth information. To address this, this study focus on the depth cues contained in underwater polarization imaging, integrates polarization features into the binocular imaging framework, and improves the accuracy and stability of depth estimation in turbid water, filling the research gap of polarization in underwater three-dimensional imaging utilizing polarization imaging technology.
