In this paper, we present a framework to represent the face of any individual, dealing with identity and expression variation and some applications of this model. A 3D morphable model (3DMM) is a generative method capable to reconstruct the 3D shape of human faces from a small set of coefficients. It is used in many applications such as identity or expression recognition, 3D scans processing or face animation. Such a 3D face generative model is used to build any face with expression as linear combination of deformations. In order to determine these basis deformations and with the hypothesis of Gaussian distribution of the 3D human faces space, a PCA is computed on a registered dataset. The registration of training 3D faces is first achieved so that it is robust to expression. To build the 3DMM with identities variations separated from expressions variations, two PCA are computed. Using this morphable model, a new face can be represented as a linear combination of these principal modes of variations. We show in this paper that such a morphable model can be used as shape prior in many applications like denoising or occlusion recuperation tool.