Abstract:
We present a region-based robust 3D face
recognition approach which is robust to facial
expressions, illumination changes and
occlusions. Facial surface is often deformed by
expressions. Generally, the mouth is the most
affected by expressions, whereas the nose is the
least affected and the most static region. For this
reason, we have concentrated on locating the
nose tip and segmenting the nose region. Our
method can be grouped into two types: The
surface-based approach which uses curvature
information of the face and the statistical-based
approach which uses subspace analysis. A new
algorithm based on the combination of these two
types of approaches is presented in this paper.
The algorithm extracts the curvature information
of the nose region from range image first, by
decomposing into maximum and minimum
curvature, and then applies PCA (Principal
Component Analysis) to reduce the dimension of
feature space, respectively. The two features are
fused using the sum rule. Our results show that
the utilization of the cropped nose region
increases the recognition accuracy up to 96.1
percent, where a subset taken from GavabDB
database is used to make evaluations.
