3D shape and pose estimation of face images using the nonlinear least-squares model

In this paper, we propose an efficient algorithm to reconstruct the 3D structure of a human face from one or a number of its 2D images with different poses. In our algorithm, the nonlinear least-squares model is employed to estimate the 3D structure of the face and the respective poses of the 2D face images concerned by means of the similarity transform. Furthermore, different optimization schemes are presented with regard to the accuracy levels and the training time required. Our algorithm also embeds the symmetrical property of human faces and the regularization term based on linear correlation into the optimization procedure so as to alleviate the sensitivities arising from changes in poses and improve the estimation accuracy of the 3D structure. Experimental results demonstrate the feasibility and efficiency of the proposed methods.