The invention puts forward a face fitting method based on a three-dimensional deformation model. The method comprises the main contents of the three-dimensional deformation model, scaled orthographicprojection, a perspective camera model and a shape from a sign. The method comprises the process that: firstly, from a shape parameter, determining the vertex position of the three-dimensional deformation model, wherein the shape is described by a linear subspace model in data via principal component analysis; then, using an axial vector to carry out parameterization on a rotation matrix, enablingan interval space to be constant, and simulating calibration lacking information; finally, assuming that the sign corresponds to the vertex in the deformation model, finding a shape, a gesture and acamera parameter, and minimizing the sum of squared distances on all signs; and importing target functions under an orthographic and perspective situation, and expressing the target functions as a separable nonlinear least square problem. By use of the method, the model is fit to server as the separable nonlinear least square problem, one series of problems brought by light, shape shielding and the like can be solved, the calibration lacking information is effectively simulated, and face fitting is quick and accurate.