Polarization dehazing method based on separating and iterative optimizing airlight from the frequency domain for different concentrations of haze

Polarization technology has been widely used in imaging through a scattering medium. However, the existing polarization dehazing methods are unstable because they require manual selections of polarization correction parameters. In addition, most of them only focus on the spatial domain without utilizing the frequency domain features, so their dehazing performances are insufficient. In this paper, we propose a polarization dehazing method based on separating and iterative optimizing airlight from the frequency domain. By separating the low-frequency sub-bands of polarization images and refining them as the airlight at three states, we calculated the Stokes parameters of airlight and obtained the preliminary dehazed image. We also propose an iterative optimization approach between the high-frequency sub-band of the dehazed image and airlight to effectively improve the dehazing performance. As a by-product, we introduce our real-world polarization datasets collected in different concentrations of haze. Both the qualitative and quantitative experiments show that our method is effective and robust in different concentrations of haze.