Abstract: Image texture analysis theory and its application in satellite imagery are discussed in this paper. Texture features of cloud imagery under different neighbourhood are extracted with the SAR random field texture model. The best neighbourhood set is choosed by using pattern recognition technique. Fractal theory is applied to the satellite imagery for its advantages in describing self-similarity of high speed spinning vortex structure. Four kinds of estimation algorithms using the theory are implemented and experimental results show that information dimension is the best fractal feature to describe typhoon. A novel texture recognition model-SARF model is proposed on the basis of much study stated above. The results indicate that SARF model can not only describe the total texture features of typhoon, but also help to reach a high precision and high speed in classification. The model can contribute much to the analyzing and processing of cloud imagery and hence to the development of meteorological science.