Automatic cloud and cloud-shadow detection and restoration on satellite images

• Main Authors: Ching-Kuo Yeh, 葉精國
• Other Authors: Victor Jung-Der Tsai
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/43863050964240722352

碩士 === 國立中興大學 === 土木工程學系所 === 94 === Satellite remote sensing has been widely used in many applications in natural science, resource exploitation and management, and environmental monitoring. However, objects within cloud shadows reflect little radiance on satellite images because the incident luminance is occluded by the clouds. As results, clouds and cast shadows may cause loss of feature information, false tone (gray or color), and failure in image interpretation within the shadow area. This research develops automatic procedures and image processing interface and modules for detecting and restoring the clouds and cloud shadows in satellite images. The color transformation technique is applied in three-band multispectral images in obtaining the intensity component as the one-band panchromatic image. The histogram of the intensity component is then used in obtaining the threshold for the segmentation of clouds as well as cloud shadows in the image, followed by noise removal by applying morphological operation and region growing approach. A cloud-free mosaic can then be produced by registering multi-temporal images and replacing the image patches of cloud. Meanwhile, the histogram matching approach is applied in restoring the brightness of the pixels within cloud shadows with respect to those in the nearby non-shadow area. Thus, the radiometric information of the ground objects within the cloud shadows is restored in the cloud and cloud shadow free images. Experimental results in manipulating IKONOS, Landsat-5, and SPOT-4 images prove the effectiveness and performance of the proposed approaches in handling clouds and cloud shadows in satellite images.