Adaptive partition based underwater polarization image restoration method for complex objects

Underwater imaging methods based on scattering models usually require the estimation of the target's polarization information. Targets in real environments exhibit complex polarization characteristics, and these characteristics are also related to the material properties and structures. In a single scene, both high-polarization and low-polarization targets may coexist, and a single target may also contain both high-polarization and low-polarization regions. Segmenting the target areas and computing for each region can effectively enhance the quality of the image's reconstruction. In this paper, we proposes an adaptive partition-based underwater polarization imaging method for complex objects. This method enables adaptive partition calculations for target images with complex polarization characteristics. It utilizes an image contribution operator to describe the contribution of regions to the quality of image recovery. The image contribution operator comprises a region size operator and an error control operator, which characterize the proportion of the partition and the polarization light component. By using the numerical value of the image contribution operator, the target image can be adaptively partitioned, allowing for individual estimation of the target's reflection light polarization for each region. This method addresses the issue of poor image recovery for targets with complex polarization characteristics. Experimental results from various underwater scenarios show that this method can achieve good recovery results for complex targets and demonstrate robustness in different levels of turbid environments.