Face recognition of identity authentication may be malicious fraud by photo, video, and other media. After analyzing the characteristics of non-linear changes in face image, we propose a new method of face liveness detection based on image processing according to the changes of edges in photo face and real face. The changes of face edges are described by the improved histogram of oriented gradient. Combining with the statistics characteristics of the positive and negative samples, the support vector machine classifier is trained to detect the liveness face. Finally, the new method is validated by the public NUAA human face database. The results show that the accuracy rate of the proposed method on true and false face detection reaches 97%.