Multi-representation Based Virtual Samples for Image Classification

Zhang Guiying,Zhao Yong,Xiang Han

DOI: https://doi.org/10.1109/CICN.2016.68

2016-01-01

Abstract:Sparse representation, which represents the test sample as a linear combination of the whole training samples, achieved great success in face recognition. It can obtain a good performance if there exist enough training samples. However, the number of face images of a subject is usually limited in real face recognition systems. In this paper, in order to obtain more representations of a face, we propose a novel method that applies the singular value decomposition (SVD) to produce virtual images of original images. Furthermore, we integrate the virtual samples and its original samples, which allows more information of the same class object to be available, so better performance can be achieved. Experiments on the most widely used and challenging benchmark datasets demonstrate that our method can obtain better accuracy and robustness in comparison with previous methods.