TOP  >  Thesis/Dissertation  >  {Modeling of face shape deformation based on displacements of facial feature points}

{Modeling of face shape deformation based on displacements of facial feature points}

{Modeling of face shape deformation based on displacements of facial feature points 3D face animation has been used in making films, games, and so on. Although 3D scanning is useful to reconstruct realistic but static 3D face model, it is not easy to produce 3D face animation by manually deforming such model corresponding to changes in facial expression. In this study, we aim to reconstruct deformation model of 3D face which represents whole facial shape with respect to the configuration of control points. We aim to clarify the necessary and sufficient set of control points to reconstruct/ identify the deformation of whole facial shape. In previous study, Bickel et al. reconstructed 3D face deformation model in which whole face surface is deformed by manipulating 89 control points sampled in whole face surface. On the other hand, Facial Action Coding System (FACS) named the deformations in facial parts based on muscles as Action Unit (AU), and suggested that whole face deformations can be represented by the combination of AU. Based on this suggestion, some of the control points in Bickel”Ēs work are not involved with AU, and these control points are displaced according to other control points involved with AU. In this study, we select based on displacements of facial feature points involved in AU as the control points. We observe 171 of optical markers put on all over the face and obtain the displacements of the markers in various expressions. For reconstructing 3D face deformation model, we conducted regression analysis in which explanatory variables are the displacements of facial feature points, and objective variables are that of other points. In experiment, we evaluated the deformation models by comparing estimated deformations and observed displacements of the markers. As a result, we demonstrated that our model can express the deformation of facial shape by 21 of control points. }