Derivation of 3D cloud animation from geostationary satellite images

Large-scale cloud animation is crucial to TV weather presentation, weather observer training and video products. In this paper, a physically based system is presented for the derivation of time-varying 3D clouds from geostationary satellite images. Cloud properties are derived from a set of meteorological models while the clouds are rendered by graphics models, the proposed method thus presents a new modeling methodology, which integrates the reality of the data with the realistic visual feeling. In particular, image pixels are first classified into cloud-free, water cloud, ice cloud, thin cirrus cloud in terms of their spectral signature. Then, cloud top surface, cloud bottom surface and cloud extinction are generated by applying different combinations of images. Finally, clouds are rendered under various light directions or view directions. The results have indicated that the proposed method can yield a realistic and approximately valid clouds with similar appearance to those in the input satellite images.