Research on Polarization Defogging Imaging Algorithm for Long-Distance Targets Based on Improved Semi-Inverse Principle

To address problems such as inaccurate atmospheric light estimation, color distortion, and insufficient preservation of edge details in traditional dehazing algorithms under haze conditions, this paper proposes a color polarization dehazing algorithm based on an improved semi-inverse principle. First, the semi-inverse principle is deeply integrated with the physical model of polarization imaging. By constructing a semi-inverse transformation model in the polarization space, pixel-level adaptive estimation of atmospheric light parameters is achieved. Second, median filtering is used to estimate the atmospheric light intensity to suppress abnormal highlight noise, and the lower bound of transmittance is appropriately raised to preserve weak signals from distant objects. Then, the transmittance is estimated based on the relationship between the polarization difference within a sky mask and the degree of polarization of atmospheric light. Guided filtering is further employed to refine the transmittance map, and an enhancement factor and gamma correction are introduced to compensate for brightness loss and improve visibility in dark areas. Finally, a local detail enhancement module is used to restore details in hazy regions, thereby improving the overall contrast of the image. Experimental results on real hazy images demonstrate that the proposed algorithm outperforms classical polarization-based dehazing methods and mainstream single-image dehazing algorithms in both subjective visual quality and objective metrics. In particular, the restoration capability for distant targets is significantly enhanced, effectively solving the problem of distant target disappearance.