Atmospheric Correction for Polarimetric Images Based on Spectral Segregation

Abstract

In hazy weather, light's penetration power is wavelength related, the longer wavelength, the less attenuation. Although traditional polarimetric image-dehazing algorithms have demonstrated their ability in enhancing grayscale images, but their ignorance of the spectral difference will lead to serious color distortion when utilizing these algorithms for color images. To conquer that problem, we propose a new method base on spectral segregation. 15 spectral bands are selected and dehazed with the polarimetric dehazing algorithm separately to obtain the best dehazing effects. The blue, green and red channels of the dehazed image, which are acquired through image fusion of the spectral bands, are adjusted with different coefficients to correct the color distortion. 10 infrared bands are added to the short-wavelength channels to enhance the details of the objects especially the trees. Experiment and data analysis demonstrate the effectiveness of our method in increasing visibility and preserving color information. The amount of color distortion can be reduced by 89.6% compared with the polarimetric image-dehazing algorithm without spectral segregation.