An automatic 3D face mesh transform based on point matching and its application

In order to achieve two 3D face mesh automatic transform, two 3D face with different topological structures are regarded as two point sets and then an automatic mesh transform method based on point matching is proposed. Firstly iterative closest point is used to align two 3D face vertex sets, then the TPS(thin-plate spline) and deterministic annealing algorithm can be used for point registration to implement automatic mesh transform, at last Kriging interpolation is applied to enhance similarity between resulted mesh and source mesh. In the end, we implement the system in VC++6.0 and OpenGL environments, and transform the default face mesh to a realistic image in Second Life using the system we established.