An unsupervised competitive neural network for efficient recognition of facial images is proposed. The proposed unsupervised competitive neural network, called centroid neural network with Chi square distance measure (CNN-χ 2 ), employs the Chi square measure as its distance measure and utilizes the local binary pattern (LBP) as an effective feature extraction tool for image data. The proposed CNN-χ 2 is applied to a face recognition problem on the Yale face database. The results are compared with those of well-known approaches including KFD (Kernel Fisher Discriminant based on eigenfaces), RDA (Regularized Discriminant Analysis), and Sobel faces combined with 2DPCA (two dimensional Principle Component Analysis). The evaluated results demonstrate that the proposed CNN-χ 2 algorithm outperforms recent state-of-art algorithms in terms of recognition rate.