A Novel Approach of 3D Face Reconstruction Using Ellipse Fitting

The problem of reconstructing the 3D face using one image captured from a viewpoint arises in many fields of science and engineering, this paper propose the novel approach for reconstructing the 3D face by using ellipse fitting technique from only one 2.5D image. The 2.5D face image is captured from any viewpoint between -45 degree and +45 degree. The cross section across the head is fitted by ellipse, and then the semimajor axis vectors are corrected to the nose tip and ridge by the nose ridge detection algorithm, the nose tip and ridge are localized. The mirror of real face is computed from the symmetry plane passing the nose ridge, the nose tip and the point on the center line of face. Finally, 3D face is reconstructed using the real face and its mirror face. This proposed method has the advantages: 1) the only one 2.5D image from a viewpoint between -45 degree and +45 degree is used; 2) the generic face model is not required in the system. The 135 3D virtual face samples were reconstructed in the experiment. The results show the feasibility of the proposed method.