In this paper, a novel approach, based on combining 3D Morphable Model (3DMM) to shape vector and texture variance, is proposed for face recognition in video (called 3DMM-S-TV). In detail, the system fits a face video to a 3DMM, then utilizes shape fitting coefficients and texture info to recognize face. For this purpose: (1) apply 3DMM to reconstruct 3D face; (2) form a shape vector to present each face in video; (3) calculate texture variance of each face; (4) use shape vector to estimate a face gallery in training data similar to faces in test data; (5) use minimum texture variance to identify objects from the gallery. Proposed methods are evaluated on two face video databases (YouTube Celebrities and FAMED).

face recognition using 3DMM, 3D face recognition.