In this work, a method for automatic construction of a 2D Active Appearance Model (AAM) of human faces from a 3D Morphable Model (3DMM) is developed. Landmark points for the AAM are selected on the 3D surface of a reference face and projected to 2D. While this task is straight forward for static feature points such as eyes or nose, it is more complicated for facial contours since they are a function of the pose angle. An algorithm for automatic contour extraction is therefore devised. The model is then fitted using the Inverse Compositional Image Alignment (ICIA) algorithm, a Gauss Newton approach similar to the Lucas Kanade algorithm but much more efficient since most of the iteartion computation that has to be done in Lucas Kanade (such as the computation of the Jacobian and the Hessian) can be precomputed. (less info)