Facial Expression Animation Generation Based On Blendshape

Facial expression animation is one of the hotspots and difficulties in the fields of computer vision,computer animation and virtual reality.During the last two decades of progressive development of facial expression animation,many scholars have carried out plenty of researches in related areas and have gained a wealth of theoretical and practical achievements.This thesis focus on the studies of performance-driven facial expression animation,including data capture,expression subspace build,and automatic generation of expression-based blendshapes.Specific research contents and innovations are summarized as follows:1.A method constructing facial expression subspace based on tensor is proposed.The basic idea of this method is to find such an expression subspace which can be all covered through core tenor,and to fit expressions by this core tenor.Besides,because more than one face and expression data are used in the process of establishing face expression subspace based on tenor,this method embodies detail features of data better than PCA and is better than PCA on the point of fitting accuracy.Experiments show that the ability of spatial generalization of tensor-based expression subspace is better and can be better suited to fit different people's facial expressions which are not existed in the expression training examples.2.An automatic generation method of example-based blendshapes is proposed.This method applies the technology of deformation transfer to achieve the coarse-grained mapping of target model from source model,and then to set up restraints of gradient space for target model and optimize expression blendshapes,and at last an alternating iterative method is used to optimize the topology of the target model.The experiment shows that our methods is feasible,and can enhance the authenticity of the models,and can generate blendshape models with high quality and individuality.3.This thesis designs and realizes the performance-driven facial expression animation system based on a combination of Kinect camera and OGRE.The system has a lot of advantages,such as simple equipment,low-cost,high precision,real-time preview of expression animation,and independent operation by a single person.