Dense surface point distribution models of the human face

In this paper we show how a dense surface model of the human face can be built from a population of examples. A technique that combines active shape models (ASMs) with iterative closest point (ICP) can be used to fit the model to new faces. The model is built by aligning the surfaces using a sparse set of hand-placed landmarks, then using thin-plate spline warping to make a dense correspondence with a base mesh. All of the mesh vertices are then used as landmarks to build a 3D point distribution model. The dense surface point distribution model is more sensitive than the landmark model to correlated facial characteristics such as gender, age and the presence of congenital abnormalities.