Facial Reconstruction and Alignment Using Photometric Stereo and Surface Fitting
This paper presents a novel 3D face shape capture device suitable for practical face recognition applications. A new surface fitting based face alignment algorithm is then presented to normalise the pose in preparation for recognition. The 3D data capture consists of a photometric stereo rig capable of acquiring four images, each with a different light source direction, in just 15ms. This high-speed data acquisition process allows all images to be taken without significant movement between images, a previously highly restrictive disadvantage of photometric stereo. The alignment algorithm is based on fitting bivariate polynomials to the reconstructed faces and calculating the pitch, roll and yaw from the resulting polynomial parameters. Successful experiments are performed on a range of faces and pose variations.
Unable to display preview. Download preview PDF.
- 1.Zhao, W., Chellappa, R. (eds.): Face Processing: Advanced Modeling and Methods. Elsevier, Amsterdam (2006)Google Scholar
- 5.Kee, S.C., Lee, K.M., Lee, S.U.: Illumination invariant face recognition using photometric stereo. In: IEICE Trans. Inf. and Syst., vol. E83-D, pp. 1466–1474 (2000)Google Scholar
- 9.Lienhart, R., Maydt, J.: An extended set of Haar-like features for rapid object detection. In: IEEE ICIP, pp. 900–903 (2002)Google Scholar
- 10.Forsyth, D.A., Ponce, J.: Computer Vision, A Modern Approach. Prentice-Hall, Upper Saddle River (2003)Google Scholar
- 13.Watt, A.H., Policarpo, F.: The Computer Image. Addison-Wesley, Reading (1998)Google Scholar
- 14.http://www.3dmd.com/3dmdface.html (February 26, 2009)