This paper puts forward a novel approach for restoration of low‐rank information fromRS images based on matrix completion ,and carried out some denoising and inpainting experiments using the singular value thresholding operator through determinate sampling and warm starting .The results indicate that the effect of our method is dominant when addressing the recovery of missing information caused by polluting and sheltering ,and moreover ,the approach can preserve details and textures in the images and make structures coherent while restoring original images ,which shows great potential in the application of RS‐image denoising and thick clouds removal .Especi ally when images possess low‐rank characteristics such as simil ar structures and regul ar textures ,the performance of the approach proposed will be better .%提出一种基于矩阵填充的遥感低秩信息复原方法，通过“确定性采样”与“热启动技术”，利用奇异值阈值迭代收缩算子进行了椒盐噪声去除与去厚云修复试验。试验表明，本文方法对于因污染或遮挡等原因造成的信息缺损问题的复原效果占优，其在信息复原的同时能较好地保留细节纹理信息并保持图像结构的连贯性。此法可用于遥感影像椒盐类孤立的点状噪声去除与厚云修复复原中，尤其是当影像矩阵具备区域结构内容相似性及纹理规则等低秩特征时，这种复原效果更佳。
展开▼
