K.-P. Min and J.-C. Chun (Korea)
TSL color model, Anatomy of the head, Fiducial featurepoints, Facial modeling
In this paper, we present an automatic approach to model animated 3D faces. In general the 3D face model is constructed by using the input of data from 3D digitizers, capturing geometry from sliced models, or a scanning laser digitizer. However, this face model doesn't contain muscle information, so it cannot be directly applied to animation. For automatic 3D face modeling we utilize Parke's facial model that is used as a generic base for acquiring muscle information and then deform a generic face mesh to fit the particular geometry of the face. To acquire range data we make use of color images. The tint saturation color information and the proper proportion of a human head is used to automatically detect the face and the external correspondence facial features in the two views of a face. We refine a generic face model by using these features and the depth information that is acquired by the analysis of a set of fiducial points in the face. In this way, we have been able to create the realistic and animated facial model of individuals.
Important Links:
Go Back
