This paper describes a method for precisely estimating the surface properties of art paintings, including surface normal, surface spectral reflectance, and reflection model parameters. We use a multi-band camera system with six spectral channels. The surface of a painting is observed under different illumination directions by the camera. We use the Torrance- Sparrow model as a reflection model. The surface normal is estimated on the different illumination observations by the photometric stereo. The spectral reflectance is estimated on the multi-band data by the Wiener method. The model parameters are estimated on the statistical distribution of the specular component. Finally, all estimates are combined for rendering various appearances of the painting under different illumination and viewing conditions.