Owing to the development of smart cities, intelligent transportation, and the Skynet project, increased attention is being paid to the decoupling technology used in the transmission and reflection of the overlapping images of car windows, which is essential for checkpoint registration, driver monitoring, fugitive trailing, and military counter-terrorism. In this study, a light transmission and reflection model based on polarization imaging technology is established and a decoupling method for transmitting and reflecting light from car windows is proposed to overcome the drawbacks of traditional imaging technology such as the reliance on intensity information and susceptibility to disturbance. The results conducted under the simulated indoor and realistic outdoor scenes show that the image information entropies of the de-reflected images of a saloon car and passenger car model are 14.3% and 9.8% higher than those of the original intensity images, respectively. Moreover, the image information entropy of the de-reflected image of a real vehicle is 2.7% lower than that of the original intensity image because the reflection contains a large quantity of environmental information. Additionally, the region contrast of these de-reflected images is improved by 40.1%, 117.5%, and 237.8%, respectively, compared with those of the original intensity images. Therefore, the relationship between the decoupling quality of the overlapping image and geometric factors of the model is studied, and a conclusion aimed at providing a reference for the installment height of the camera and position of the vehicle stop line is provided.