Virtual samples construction using image-block-stretching for face recognition

Abstract

Face recognition encounters the problem that multiple samples of the same object may be very different owing to the deformation of appearances. To synthesizing reasonable virtual samples is a good way to solve it. In this paper, we introduce the idea of image-block-stretching to generate virtual images for deformable faces. It allows the neighbored image blocks to be stretching randomly to reflect possible variations of the appearance of faces. We demonstrate that virtual images obtained using image-block-stretching and original images are complementary in representing faces. Extensive classification experiments on face databases show that the proposed virtual image scheme is very competent and can be combined with a number of classifiers, such as the sparse representation classification, to achieve surprising accuracy improvement.