A Fast Removal Method of Thin Cloud/Haze Cover for Optical Remote Sensing Images Based On Multi-fractal

Because of the presence of cloud, the quality and application of data obtained in visible light and infrared bands are affected when remote sensing images are generated. This paper proposes an automatic selection method with the best cut-off frequency to remove the effect of thin cloud in remote sensing images quickly and efficiently. Based on the homomorphic filtering method of the simple thin cloud imaging model and the self-similarity of the spatial form of thin cloud and haze, this paper uses the multi-fractal technology and the filtering technology of the S-A model (power spectrum-area model) to determine the filtering radius automatically. The experiments of removing thin cloud and haze for remote sensing images show that the filtering technology of the S-A model is related to the spatial form of image contents directly. Compared with usual filtering technologies which choose frequency as radius, this method can not only achieve rapid and automatic filtering of remote sensing images, but also remove the effect of thin cloud in remote sensing images effectively.