Detecting edges has become more difficult in image processing since images may contain areas with different degrees of noise.Many traditional methods for edge detection are sensitive to noise,so that the image edge can not be detected clearly. This article presents a noisy image edge detection method  based on zero crossing. This method first smoothes the image and calculates gradient on the basis of edge detection operator in reference［1］,then calculates the second derivative of the gradient image by the new recursive operator,and finds out the zero crossing points by rows and columns respectively,and finally merges all the zero crossing points and gets the image edge.The experimental results show that this method not only gives good edge detection effect,but also spends less computation and time by the recursive implementation of filtering operators.