Marker-less Facial Expression Capture And Transfer Algorithm

Marker-less facial expression capturing technique is a new direction of the facialcapturing technology, and different from the mature marked facial capturing techniqueat present, and has greater difficulty in the data acquisition. So there is more researchspace. This technology combines image processing, computer vision, computer graphicsand many other disciplines and is the reaserch hotspot in recent years. It not only hastheoretical significance but also has wide market value, can be used in many field, suchas medical analysis, studio entertainment, game production and so on. This paperstudies several key issues in marker-less facial expression capture and makes someachievements.In this paper, there is a more detailed introduction and analysis about the faicialexpression capture at the present stage and has deep study on the key issue of themarker-less facial expression capture. Firstly, building a novel model for describing thefacial expression changement, the model includes local features and global mesh. Thelocal features are from identifying the face by ASM (Active Shape Model) algorithm,and then the global mesh is generated evenly according to the position of featurepoints. This model covers most part of the face and can describe the most facialexpression accurately. Secondly, this paper proposed a new method to capture facialexpression under the basis of the face model. The optical flow tracking is used to trackthe feature points with particle filter for stablizing the tracking results. The displacementof the feature points can drive the overall mesh to change as the initial value of the meshtracking. The mesh deformation algorithm is used for driving the mesh and the linearand nonlinear methods combined for the mesh tracking to realize the expression capture.Finally, the data from the expression capture is used to drive different face models,using different meshod for the models of different dimensions, and then the facialanimation has been reconstructed. The animation effect is good for different2D cartoonor3D virtual face models to which the expression is mapped.In this paper, there builds a prototype system for expriment and uses the C++language to realize capture algorithm for sofrware. Many groups of experimental dataand the better result can verify the feasibility of the marker-less facial expressioncapture algorithm based on mesh tracking.