Constructing Faces using Active Appearance Models and Evaluating the Similarity to the Original Image Data
Andre Stormer, Jan Stadermann · 2004
Active Appearance Models (AAMs) can be used for interpreting face images and image sequences. AAMs combine a statistical shape model and a model of grey-level appearance. They also contain an iterative matching scheme for image interpretation, which only needs an initial estimate of the position and size of the face and results in a set of parameters describing the matched face. In this paper we describe, how to build up the shape model, the grey-level model and how to combine them to an AAM. Then we show, how to construct a search algorithm to derive the model parameters for a given face and investigate how well these parameters can be used to descibe known and unknown face images, measuring the similarity between the model-built faces and the original image data.