Abstract: The principle and implementation method of surface reconstruction by analyzing the polarization images of reflected light to reconstruct the surface shape of a transparent object are presented. The light reflected from an object has some polarization characteristics, and different polarization characteristics correspond to different shapes and reflectance characteristics of objects. The surface shape of an object can be reconstructed by analyzing the polarization distribution of the reflected light from the object surface. The functional relation between polarization ratio and surface normal was derived according to the Fresnel formula. The surface shape of an object under test was reconstructed by processing the polarization gray-images taken by CCD camera. Experiment results show that the proposed method is feasible and efficient.