Polarimetric dehazing utilizing Vis-NIR spectral image fusion

Polarimetric dehazing methods have good ability in restoring images contaminated by haze. But noise and color distortion come along with the dehazing process, and those methods lack the ability to recover targets of long-distance. Our previous study has shown that the penetration ability of light is related with wavelength, light of long wavelength, especially in the infrared band, receives less attenuation from haze. Based on that fact, we decompose the dehazing process into 16 different spectral bands to achieve the optimized dehazing performance. An infrared band is added to enhance the dehazing performance in short-wavelength channels, thus increases the range of visibility and corrects the color distortion. The dehazed image is reconstructed through image fusion of the dehazed bands. Experimental analyses reveal enhanced details and reduced noise of dehazed images. Particularly, our method works well in retaining the color information of haze contaminated images. The amount of blueshift can be reduced by 71% at least.