A Statistical Two-step Method for 3D Face Reconstruction from a Single Image

To make computer recognize well about the facial shape, we propose a Two-step face modeling (TSFM) scheme to make use of prior knowledge of facial shape learnt from a large scale 3D face database. At first, the depth of features is estimated. Then we apply the estimated 3D features to a novel dynamic components deformation model to build a more precise facial shape for a given image. Test results on synthesized images with the known ground truth show that the TSFM scheme is both robust and accurate in facial shape reconstruction.