• A realistic and diversified training dataset (POL) is constructed by optical modeling for reflection obstructed image. • A progressive polarization based reflection removal network (P 2 R 2 Net) with up-to-date neural network units is proposed. • The improvement of the proposed method for real-world images is verified by comparisons with the state-of-the-art methods. • Robustness analysis on reflection-tainted synthesis images suggests our method possesses excellent generalization ability. The reflection effect is unavoidable when taking photos through glasses or other transparent materials, which introduces undesired information into pictures. Hence, removing the influence of reflection becomes a key problem in computer vision. One of the main obstacles of recent learning based approaches is the lacking of realistic training data. To address this issue, we introduce a new dataset synthesis method as well as a novel neural network architecture for single image reflection removal. First, we make use of the polarization characteristics of light into the synthesis of datasets, so as to obtain more realistic and diversified training dataset POL. Then, we design a novel Progressive Polarization based Reflection Removal Network (P 2 R 2 Net), which preliminary estimates the coarse background layer to guide the final reflection removal. We demonstrate that our method performs better than the state-of-the-art single image reflection removal methods through quantitative and qualitative experimental comparisons. Specifically, the average PSNR of our restored images selected from three representative benchmark datesets: "Real20", " SI R 2 " and "Nature" is improved at least 0.49 compared with existing methods and reaches to 24.52. [ABSTRACT FROM AUTHOR]