Real-Time Domain Transform Recursive Filtering Dehazing Method Based on Polarization Image

Compared with the visible image restoration details, the polarization image is more abundant with the polarization characteristics to restore the target details. In order to overcome the problem that the current polarization-based dehazing algorithms, which the model parameters is susceptible to high-brightness objects and the noise of the restored image is more, we propose a universal polarization-based dehazing algorithm based on domain transform recursive filtering. On this basis, the quadtree spatial indexing method, the dark channel prior and the domain transform recursive filtering are researched to effectively estimate the model parameters. Then, the limitations of the parameters depending on the sky are broken through. Finally, according to the physical polarization-based dehazing model, the haze-free image can be well restored. Our experimental results show that the proposed approach not only enriches more texture details and reduce the noise as well, but also improves the time complexity of the algorithm.