Automatic Construction of Active Appearance Models as a Face Synthesis Problem

This paper describes the principle of automatically constructing an active appearance models as a face synthesis problem. An algorithm is presented to find the optimal AAM model rapidly and improve the performance of AAM on the face synthesis problem. The feasibility of method proposed here is validated by some of the face synthesis examples. Experimental results show that our approach not only ensure the classification accuracy of image interpretation but also improve the performance of convergence with other standard methods.