Statistical Face Recognition and Intruder Detection Via a k-means Iterative Algorithm: a Resampling Approach

In this paper a face recognition algorithm based on a iterated k-means classification technique is introduced. The suggested algorithm, when compared with popular PCA (Principal Component Analysis) algorithms for face recognition more than holds its own. The presented algorithm, unlike PCA, is not a dimensional reduction algorithm, nonetheless it yields barycentric-faces which can be used to determine different types of face expressions, light conditions and pose. The accuracy of PCA and k-means methods has been evaluated under varying expression, illumination and pose using standard face images. Using a resampling approach we will show how the proposed technique can be used to detect intruders, i.e. pictures of objects or persons not in the database.