Estimating the shape and appearance of a three dimensional object from flat images is a challenging research topic that is still actively pursued. Among the various techniques available, Photometric Stereo (PS) is known to provide very accurate local shape recovery in terms of surface normals. In this work we propose to minimise non-convex variational models for PS that recover the depth information directly.Photometric Stereo consists in finding a depth map z that best explains all image irradiance equations (IIEs) I i = R(z; s i , ρ), for several images I i , considered under different lightings s i , with i ∈ {1, . . . , m}. The function R describes our reflectance model in terms of the depth z, the lighting s i , and the albedo ρ. We assume Lambertian reflectance, neglect shadows, and require m 3.Our approach uses a variational framework with a least-squares penalisation on the IIEs augmented with a zero-th order Tikhonov regularisation. The obtained energy (1) is non-convex and we make use of matrix differential theory and recent developments in non-convex and nonsmooth optimisation to determine good minimisers. min z,ρ