Photometric stereo using graph cut and M-estimation for a virtual tumulus in the presence of highlights and shadows

The photometric stereo method is useful for modeling the fine detail of the surface shape of an object. In this paper, we propose a photometric stereo method that uses a graph cut solution. We formulate the photometric stereo problem to the Markov random field problem, and show how to solve the problem by graph cut. The graph cut properly calculates the surface normal and automatically evades the interference of specular reflection. Finally, we show some results when our method is applied to common objects (a diffuse object and a specular object), and to cultural assets, the Segonko Tumulus and Sakurakyo Tumulus.