Single sample face recognition based on virtual images and 2DLDA

Jun Yang, Yanli Liu

Abstract: When there is only one sample per person in gallery set, the conventional face recognition methods which work with many training samples do not work well. Especially, a number of methods based on Fisher linear discrimination criterion cannot work because the within-class scatter matrix is a matrix with all elements being zero. To solve this problem, a method was proposed to get virtual sub images of one face by an image processing method. With these virtual images, the within-class scatter matrix can be evaluated and the supervised learning method such as 2D fisher linear discrimination analysis can be utilized for feature extraction. The experimental results on ORL face database show that the proposed method is efficient and it can achieve higher recognition accuracy than others.