A model for digital image computation of a driving scene during daytime fog has been implemented in order to evaluate targets' visibility. Using a semi-Monte Carlo method to solve the radiative transfer equation inside fog, we establish a physically based rendering with an accurate lighting simulation. The fog is modeled as a layer of water droplets described by a Deirmendjian size distribution, and the scattering properties of each droplet are computed using Mie theory. The boundary conditions are sky luminance and sun illuminance on top of the fog layer, and a bidirectional reflectance distribution function (BRDF) model derived from measurements on road samples. The veiling luminance inside the layer in all directions and for several altitudes is computed and allows us to evaluate the attenuation of every surface existing in the scene. As an example, the visibility of a Lambertian reflective silhouette is given. (C) 2008 Society of Photo-Optical Instrumentation Engineers.