Sentinel-2 Image Dehazing using Correlation between Visible and Infrared Bands

Several atmospheric phenomena often contaminate multispectral remote sensing images. Haze is one of the atmospheric phenomena that obscure the transparency of the scene. It affects the quality of an acquired satellite image. In this paper, we propose the dehazing method for Sentinel-2 satellite imagery. The correlation of the visible bands and infrared bands (narrow near-infrared (NIR) and short-wave infrared (SWIR) bands) are considered for constructing the linear combination formulas. The linear coefficients are determined by using the multiple linear regression of the haze-free image samples. The results show that our proposed method can be used to remove the haze, both thin and thick haze, especially the haze in the smoking area.