Photometric Stereo refers to the problem of 3D-reconstruction of a scene, given several images taken from the same viewpoint, but under different lightings. Since the early 1980s, researchers have devoted a lot of energy to improving the accuracy of this technique, through the introduction of realistic reflectance, camera, surface and lighting models. In this talk, recent advances in the field of photometric stereo with realistic, physics-based lighting models, will be presented. Clues for numerical resolution will be provided through the study of some energy minimisation approaches of photometric stereo, including matrix factorisation approaches for linear models, optimisation on discrete grids for non-convex functionals, and calculus of variations. It will be shown that, combining such robust optimisation tools with realistic model, 3D-reconstruction through photometric stereo is now almost mature enough to be considered as a real metrological technique.