3D Head Pose Estimation without Feature Tracking
FG '98 Proceedings of the 3rd. International Conference on Face & Gesture Recognition
Face recognition from one example view
ICCV '95 Proceedings of the Fifth International Conference on Computer Vision
Face Pose Discrimination Using Support Vector Machines (SVM)
ICPR '98 Proceedings of the 14th International Conference on Pattern Recognition-Volume 1 - Volume 1
| Hi-index | 0.00 |
Given an image of a face and the co-ordinates of the eyeballs' centre, we provide a method which creates the corresponding frontal face and gives a qualitative and quantitative measure of the horizontal tilt of the face, using the natural symmetry of the human head. By multiscale dynamic programming, we obtain the optical flow (2D warp) which transforms the input face into its mirror images, this mirror images being a good approximation of the same head that has undergone a 3D rotation. The frontal face is obtained by applying half of the magnitude of the optical flow, and any pose is obtained by applying a fraction of this optical flow. The nose axis, which gives a quantitative and qualitative tilt estimation, is obtained by the reverse optical flow applied on the symmetrical axis of the generated frontal face. Experimental results on 500 images and on the Surrey database confirm the performance of this system.