3D face model reconstruction based on stretching algorithm
Qinyan Zhang, Li Shi · 2012
In order to get realistic face model, face model reconstruction is a very important process, we need a way to map the two-dimensional face image coordinate to standard face model, so we propose a general face model reconstruction process based on stretching algorithm which including three stage. In the pose recovery stage, the pose recovery formulation is based on the non-linear least squares structure-from-motion algorithm. In the face partial point adjustment stage, which we define the stretching coefficient to adjust the feature point and non-feature point separately. After we make the face point in the right position for the model, from scattered data interpolation stage which based on the radial bases functions we can get a general function to calculate accurate 3D face model coordinates. Compared with other algorithm, the stretching algorithm can improve the accuracy of a three-dimensional map from the standard face model to a specific human face.