Modeling large scale clouds from satellite images

Visualization of satellite cloud images plays an important role in atmosphere analysis and weather prediction. However, reconstruction of meaningful 3D clouds is a challenging problem due to the 2D nature of the input data. In this paper, we present a new method for modeling large scale clouds based on cloud property retrieval theory. In contrast with previous methods, the proposed one is more physical and focuses on the geometric structures of clouds. Image pixels are first divided into cloudlessness, water cloud, ice cloud, and thin cirrus cloud in terms of spectral characteristics. Then, the top height, geometry thickness and extinction volume of the cloud are generated by applying different spectral combinations of images. Finally, clouds are rendered with various light directions or view directions. The results show that the proposed method can not only yield realistic clouds, but also approximate actual clouds, thus being useful for time critical applications.