Pose-Invariant Face Recognition Using A Single 3D Reference Model

Approaches for cross-pose face recognition can be categorized into 2D image based and 3D model based. However, only a small number of 3D model based approaches are reported although 3D information is considered crucial for cross-pose analysis. Extended from a latest face reconstruction method using a single 3D reference model, this study focuses on using the reconstructed 3D face for recognition across poses. Given a 2D face image of frontal pose, one can reconstruct its 3D model from a 3D reference face using spherical harmonics to approximate the reflectance across the given face. Experiments on the PIE database show that addition of local correspondences for pose alignment, multiple reference models considered in the reconstruction phase, and the addition of a profile pose to the gallery set can make the performance competitive to the stateof-the-art.