Abstract: This article presents a edge detection method, which is based on the dual-threshold nonlinear derivative operator.Firstly, gray image is used to calculate the right and left derivative.Secondly, dual-threshold is used to injust the right and left derivative to remain the meaning edge information.At last, the two derivatives are merged to get the image gradient.can control the ability of reducing noise;can ensure detection of the one pixel width edge lines;and the nonlinear derivative scheme can solve the problem of delocalization.By comparing and analying it with the traditional discrete gradient operators in the experimental demonstrations, we find that this operator is not only with advantages of sinplicity, higer accuracy of detection, but also get better edge image and higer SNR in the case of no any smootning.