Based on the hourly infrared cloud images from the geostationary meteorological satellite (FY2) and the enhanced precipitation observation from the regional mesoscale surface automatic weather observation network, through the parameterization of cloud images, the relationship between cloud parameters and precipitation rate of different categories are analyzed by using the dynamical correlation method. The cloud parameters include the twoband combination of longwave infrared and splitwindow, the twoband combination of longwave infrared and water vapor, and brightness temperature, etc. Through the above analysis, a dynamical estimation model of precipitation nowcasting is built based on the optimal correlation between precipitation rate and related cloud parameters. The new estimation model was applied in real time to threehour precipitation forecast. The distribution of forecasted precipitation is similar to that from observation. The precipitation centers are roughly in agreement with the observed centers, but the estimated precipitation rates are lower than the observed in high mountain areas.