A new feature extraction method based on multi-directional correlation function is proposed for improving the classification performance of halftone images in this paper. The correlation function is improved to obtain feature vectors on three directions of halftone images. Based on the definition of pixel probability, pixel entropy and entropy histogram, we introduce a decision rule to reject some invalid halftone images. We describe the details about feature extraction algorithm of halftone image. The experiment results show the proposed method is effective.