Video Face Recognition Based on the Convex Hull Model of Kernel Subspace Sample Selection