S. Zhan (PRC, Japan), T. Kurihara, N. Ono, and S. Ando (PRC)
Correlation image sensor, depth map, curvature feature, geometrical correction, eigen-face, imaging
Facial authentication is one of the hottest areas in image processing and pattern recognition. The recognition and analysis of human faces based on 3-D images and features are also an important field. This paper presents one novel method of recognizing faces in arbitrary poses based on intensity image and depth map which can be acquired simultaneously by the correlation image sensor (CIS) [3]. By using them, we implement an algorithm to convert the facial view in an unknown pose into the standard frontal pose. In this approach, we use viewpoint-invariant curvature features of the depth map as the control points for the conversion. After getting standardized frontal view posed intensity image and depth map, we employ eigenface technique to get the authentication result. The experiments show that the performance for our test datasets is very satisfactory.
Important Links:
Go Back
