A Method for Robust Nose Tip Location across Pose Variety in 3D Face Data

3D face data have the potential to improve the face recognition performance at disadvantageous conditions, such as uncontrolled illuminations, pose and expression. 3D nose tip location is the crucial step for registration and recognition of 3D face data. An automatic nose tip location and pose estimation method robust for pose changing at 3D face scans is proposed in this paper. We firstly gain some tip candidates by directional maximum and normalize the 3D scan data to front by hypothetical nose tips. Two filters respectively based on local structure of curves and PCA reconstruction curves errors are utilized to discard false tip candidates. The results of nose tip position and face recognition at BJUT 3D Database show the proposed method is efficient and robust for pose angle variety.