3D Face Recognition using Point Cloud Kernel Correlation

3D Face recognition is a widely researched problem for wick many methods have been proposed. Unfortunately, most of these methods need vast prior knowledge in the form of large 3D face training sets. Also, many methods need point correspondences, wich are hard to obtain. In this paper, we present a general 3D face recognition method based on point cloud kernel correlation and kernel density estimation for registration, which doesn't make use of a training set of faces or point correspondences and can handle noisy, unpreprocessed face scans. Initial experiments on the publicly available SHREC 2008 3D face database show good face recognition accuracy and robust behavior across face expressions.