摘要： 提出了一种基于无抽样小波变换的人机交互半自动云阴影去除算法。首先通过无抽样小波变换将图像分解成高频部分和低频部分，通过半自动人机交互方式在低频图像中将图像分割为云区域、阴影区域、清晰地物区域3 个部分，然后对各个独立云阴影区域分别进行低频信息补偿和高频信息补偿，最后通过无抽样小波反变换得到一幅过渡平滑，并对图像细节保存完好地去除了云阴影的图像。实验结果表明，该方法能够取得比传统的直方图匹配更好的效果。

关键词: 阴影检测；阴影补偿；无抽样小波变换；直方图匹配

Abstract: A semi-automated approach for cloud shadow removal in satellite images is proposed based on undecimated wavelet transform. The image is first decomposed by undecimated wavelet transform. In the low frequency image, the shadows and clouds are detected and the image is segmented into three parts: cloud regions, shadow regions and non-shadowed regions through semi-automated human-computer interaction. Then the shadow regions are matched to the non-shadowed regions in the low frequency content and the high frequency contents in the shadow regions are also compensated. At last, the image reconstructed using inverse undecimated wavelet transform would be shadow-free. The experiment results show that the method is valid and performs better than histogram match

Key words: Shadow detection; Shadow compensation; Undecimated wavelet transform; Histogram match