Firstly, according to intensity distributions of reflected light and transmitted light in perpendicular and parallel directions on surface of transparent objects, the functional relation between the components of reflected light and transmitted light and that between the polarization degrees of reflected light and transmitted light are derived with the principle of polarization orthogonal decomposition in this paper. Then, based on the imaging principle of camera, the angles between incident lights and normal vectors and the azimuth angles of the incident planes at different positions of transparent object surface are obtained by using surface normal vector to summary the distribution rules of the polarization degrees of reflected light and transmitted light at each pixel in the image. Finally, according to the polarization characteristics and correlation features between reflection and transmission, the minimum of normalized cross-correlation between reflection component and transmission component is calculated by gradient descent algorithm to realize the separation of reflected light and transmitted light, which suppresses the interference of reflected light on the surface of transparent objects, and it plays an important role in information processing and applications such as target detection and image matching in environment with complex reflected light.