Information reconstruction in the cloud removing area based on multi-temporal CHRIS images

Abstract

Thick cloud cover is a very common problem in remote sensing images. Cloud and shadow should be removed, which causes some dark holes in remote sensing images. It is difficult to predict the real value of one pixel covered by thick cloud. Multi-temporal information extracted can be used in the information reconstruction in the cloud covering area. In this study, one new method which is different from the filter processing and spatial interpolation processing is applied in information reconstruction based on multi-temporal CHRIS images. First, the threshold method and MLC are combined to completely identify the area of cloud and shadow. Then this paper utilizes the class information extracted from multitemporal images and the statistical relationship of the reflectance at the same band in different temporal image. In the end, the whole cloud covering areas in one CHRIS image are repaired and filled by the predicted reflectance values. From the reconstruction result in the cloud covering area, it is concluded that the method in this paper is rather effective to repair the information in the thick cloud covered area based on multi-temporal images.