The invention provides a method of classifying polarimetric SAR (synthetic aperture radar) images based on scattering entropy and three-component decomposed plane, and relates to the technical field of radar image processing. The method divides target ground objects into 9 classes initially according to scattering entropies and Freeman three-component, calculates a clustering center according to the 9 classes, and clusters the target ground objects to expected quantity according to Wishart distance; the ground objects are divided into high-entropy scattered ground objects, medium-entropy scattered ground objects and low-entropy scattered ground objects according to the scattering entropy H, and the 3 classes are then divided into 9 classes of ground objects according to surface scattering, even scattering and volume scattering, and initial classes are further divided by Wishart classifier. More accurate classification may be provided for polarimetric SAR images, no approximation occurs during processing, detailed information may be well retained, and the number of mistaken classes for urban areas is decreased greatly.