In order to describe the polarized bidirectional reflectance distribution function of the paint material more accurately, a double-Gauss polarized bidirectional reflectance distribution function(BRDF) was modeled and simulated which based on the micro-facet assumption. Relative to the previous models, it contained non-Lamb and polarized diffuse reflection based on Rahman-Pinty-Verstraete function. Several experiment groups of polarized BRDF in the plane of incidence of painted surfaces, such as white, matte white, pearl white paint, and the paint on the back of a cd, were measured by the polarization multi-spectral imagery based on Liquid Crystal Variable Retarders(LCVR) at 476 nm. Then, the parameters of model were inversed step by step using the least square method from some experimental data. The inversion results were consistent under many initial conditions, which indicated the obtained parameters were the global minimum. Finally, the model parameters obtained were simulated. The simulation results prove that the simulated intensity and DOP of reflection of painted surfaces fit in the measured appropriate values very well, especially for the polarized diffuse reflection. This model can be applied in classification and recognition of material and scene simulation.