Simulation of light propagation in a dispersed medium is usually based upon continuous medium approximation, which is good when the distance between different ray hits much exceeds the particle size. But when the different ray events are not statistically independent, which violates the continuous medium approach. Current paper investigates the problem for automotive paints where the particles are thin planar flakes. We performed accurate ray optics simulation. Here subsequent light scattering events are correlated: if incident ray reached the given flake, then the probability that the reflected ray leaves the paint area without further scattering is higher than the probability of hitting a flake “on average”. If, however, the reflected ray hits the next flake while going upwards, then it will be reflected downwards and most likely hit the first flake again. After that the probability of hitting the same second flake is increased as compared to mean value. This increases the probability of uneven scattering while decreases that of even scattering. We demonstrate how this affects the total scattering and obtain some analytic estimates. We compare the bidirectional reflection function of paint surface calculated for the two models and show how the difference changes with concentration and flake size. It happens that a serious change is in the near-specular region. Some analytically derived “correction” terms can be applied to the continuous medium approach to move it towards the results of the explicit model. In some cases this improvement can be a due compromise with more expensive explicit one.
|