遥感图像云雾处于相对低频、景物处于相对高频。根据小波变换多分辨率的特点,分析云雾和景物在小波系数中的分布,得出景物信息包含于低层细节系数,而云雾信息包含于高层细节系数和近似系数。本文通过增大图像的低层细节系数,减小高层细节系数,并适当减小近似系数,达到去除云雾的目的。最后评价了图像处理结果,并与同态滤波做比较,表明本文方法是有效的。

In remote sensing images the frequency of thin cloud and mist is lower than that of sceneries. The frequency distribution characteristics of the remote sensing image influenced by the thin cloud and mist are analyzed based on the multi-resolution of the wavelet function. We conclude that the information of the sceneries are in the detail coefficients of lower levels, while the information of the cloud and mist are in those approximate and detail coefficients of higher ones. By increasing the detail coefficients of lower levels, and decreasing those of higher ones and approximate coefficients, the goal for removing the thin cloud and mist is achieved. The experimental resuhs of the proposed algorithm, being evaluated and compared with the homomorphism filtering method, are proved to be satisfying.