It is popular to extract discriminant features using Fisher linear discriminant analysis (LDA) for general pattern recognition. LDA aims to find an optimal discriminant transformation matrix, which maximizes the ratio of between-class scatter to within-class scatter. However, in case of small sample size and high dimensional data, LDA is prone to be unrealizable due to the singularity of scatter matrices. In this paper, we present a nonsingular transformation prior to performing LDA. This method is to transform general features using all eigenvectors of scatter matrix with nonzero eigenvalues. As a result, the scatter matrix of transformed features is nonsingular. Subsequently, the discriminant transformation is applied according to LDA using the new scatter matrices. The superiority of nonsingular discriminant analysis of between-class matrix comes from the shrinkage of within-class scatters and accordingly the enhancement of Fisher class separability. From the experiments on facial databases, we find that the nonsingular discriminant feature extraction achieves significant face recognition performance compared to other LDA-related methods for a wide range of sample sizes and class numbers.