A geostationary meteorological satellite plays an important role in modern weather forecast and public services. The data processing system of the geostationary meteorological satellite is a very complex system. Because of the problems of the software and hardware used in satellite and ground systems, the quality of the cloud images is often affected. A method based on the Euclidean distance algorithm is proposed. It can be used to evaluate the quality of the cloud images produced in real time. The feasibility of this algorithm in the ground system of the geostationary meteorological satellite is discussed. The comparison of the result with the historical data shows that this method can identify at least 90 % of inaccurate cloud images.