Low Dimensional Surface Parameterisation with Applications in Biometrics

This paper describes initial results from a novel low dimensional surface parameterisation approach based on a modified iterative closest point (ICP) registration process which uses vertex based principal component analysis (PCA) to incorporate a deformable element into registration process. Using this method a 3D surface is represented by a shape space vector of much smaller dimensionality than the dimensionality of the original data space vector. The proposed method is tested on both simulated 3D faces with different facial expressions and real face data. It is shown that the proposed surface representation can be potentially used as feature space for a facial expression recognition system.