Face Recognition Using Constrained Active Appearance Model

The Active Appearance Model (AAM) has been shown powerful for modeling images of deformable objects. AAM uses Principal Component Analysis (PCA) based linear subspaces to model the 2D shapes and textures of the images of a target object class. Such a representation allows AAM to represent a certain image very small number of parameters. A suitable constrained initialization is presented in matching process so as to make converge better and quicker. We demonstrate the application of the scheme to the task of face recognition and experimental results show it is effective.