Independence component analysis (ICA) is applied in Pose-varied face recognition. Discriminations between ICA and principal component analysis (PCA) in face recognition are analyzed, and independent component (IC) representation in pose-varied face is discussed in detail. Based on the method that selects a subset as the kernel for the representation by ordering the sources via the magnitude of the corresponding weights, a novel IC representation of pose-varied face based on the scale factor is proposed. Demonstration indicates that the proposed method is efficient.