Abstract: In order to obtain polarization images without true polarization images, firstly, the degree of polarization of various objects were found in experiments, and a database of degree of polarization was built. Then, objects in the infrared images were classified by means of K-means clustering algorithm. Finally, the polarization images were calculated from the clustering images. The results verify the simulation algorithm is feasible. The effect of the simulated images is good, it is easy to distinguish artificial object from natural objects. Image simulation is important for better and more convenient research of polarization properties of objects and the establishment of a database of polarization.