Nose Tip Region Detection in 3D Facial Model across Large Pose Variation and Facial Expression
Detecting nose tip location has become an important task in face
analysis. However, for a 3D face model with presence of large
rotation variation, detecting nose tip location is certainly a
challenging task. In this paper, we propose a method to detect
nose tip region in large rotation variation based on the
geometrical shape of a nose. Nose region has always been
considered as the most protuberant part of a face. Based on
convex points of face surface, we use morphological approach to
obtain nose tip region candidates consist of highest point density.
For each point of each region candidate, a signature is generated
and evaluated with trained nose tip tolerance band for matching
purpose. The region that contains the point which scores the most
is chosen as the final nose tip region. This method can handle
large rotation variation, facial expression, combination of all
rotations (yaw, pitch and roll) and large non-facial outliers.
Combination of two databases has been used; UPMFace and
GavabDB as training data set and test data set. The experimental
results show that 95.19% nose tip region over 1300 3D face
models were correctly detected.
Keywords: Nose Tip Region Detection, Morphology, 3D Face
Model, Point Signature, Tolerance Band
Download Full-Text
