基于云层系数的遥感图像去云雾算法

An algorithm for removing cloud and mist from remote sensing images based on cloud layer coefficients

讨论了数字图像小波分解过程,根据分解系数的频率分布,给出通过小波分解分离单幅遥感图像中云雾和景物的途径。探讨暗通道先验理论,分析云雾和景物暗像素值的差异。在此基础上,建立云层系数的概念,并提出一种根据云层系数从遥感图像中去除云雾的算法。即利用云层系数,估算各像素的云雾浓度,处理高层细节系数以衰减云雾,处理低层细节系数以突出云下景物。给出详细的实现步骤,进行去云雾处理实验,并评价实验效果。结果表明,去云雾效果优于小波加权算法和小波阈值算法。

The procedure of processing digital image through wavelet decomposition is analyzed. The frequency distributions of wavelet decomposition coefficients are investigated. The approach to separate the cloud and mist from the scenery by wavelet transformation is presented. The prior theory of dark channel is discussed. Differences of dark pixel values between the cloud and mist and the scenery are analyzed. On this basis, the concept of cloud layer coefficients is put for- ward. An algorithm for removing the cloud and mist from remote sensing images is proposed. That is, thickness of the cloud and mist of every pixel is estimated according to its cloud layer coefficient. The cloud and mist is attenuated by processing detail coefficients in high levels, and the covered scenery is brought out by processing those in low levels. The implementation steps are described completely. Removing cloud and mist experiments are carried out and their results are evaluated. It is proved that the algorithm is superior to the weighted wavelet algorithm and wavelet threshold algorithm.