AAM-Based Face Modeling and Landmark Selection

A basic theory of Active Appearance Model (AAM) is described and used for human face modeling. AAM is based on statistical models. The method establishes the model not only for facial shape but also for the facial texture which reflects the gray levels of face. Principal component analysis is used to establish the AAM face model. The triangulation method for texture extraction as well as the process of PCA modeling is represented in detail. The facial landmarks draw the shape of the main features of human face and the landmarks are the basis for modeling. The number and the location of landmarks may affect the human face modeling. Some experiments are done in this paper to see the influence.