Photometric Stereo for Non-Lambertian Surfaces Using Color Information

One robust method to reconstruct shape is photometric stereo (PMS), which reconstructs surface orientation using the Lambertian reflection properties of the surface material. To increase the applicability to non-Lambertian surfaces, we extend this method using a twostage process without introducing additional light sources or assuming a known micro facet distribution. In the first step, the glossy reflection is discarded, taking the dichromatic reflection model as a basis. The introduced spherical chromaticity space has been found as a good tool for deriving the matte vector. The method is applied to multi-colored objects. The simulations use spectral distributions of real data. In the second step, we apply a conventional PMS to the derived matte images.