Design And Implementation Of 3d Face Animation Based On Feature Extraction

With the advent of three-dimensional epoch in computer vision, the rapid advance of computing capacity and the flourishing development of new media industry, the 3D face modeling and animation have become one of the hottest research spots in the computer graphics research field which have derived a lot application, such as communication, man-machine interaction, multimedia, entertainment industry etc.At the beginning of this thesis,the modeling technique, deformation technique and drive technique of the three aspects of three-dimensional facial animation technology will be reviewed for proposing a unsupervised procedure of constructing and steering 3D facial model by extracting the features of some front facial images by integrating some scattered research spots.The active shape models algorithm is a statistic model based local structure extract algorithm. The statistic model is trained by lots of image samples. After one iteration finishes, the statistic model will constrain the new points matched by local structure to a reasonable feature output.Facial features searched by active shape models algorithm carry not only the scale and position of the face putting through image also texture coordinates of these feature points included. A new 3D model will be reconstructed by transforming a neutral face reference model which is coupled to the features preprocessed; A facial animation will generate by applying a geodesic distance based radial basis function to model deformation according to the MPEG-4 code that detecting the motion trend through active shape models algorithm.In the last part of this thesis, a design of efficient rendering platform which integrate computation resources of CPU and GPU balancing the computing intensity of the facial animation application is proposed.