Based on machine vision, one method used for detecting the surface defects of ceramic bowls is proposed. This method is mainly used to accomplish the edge detection of surface defects by means of Kirsch operator in combination with Canny operator. It adopts the eight direction templates from the traditional Kirsch operator to calculate the derivative convolution of every pixel point, singles out the optimal template, and determines the edge direction. The high signal-to-noise ratio, high detection accuracy and edge detail keeping of the Canny operator are combined to accomplish the detection of surface defects. Whether there is a defect or not is determined based on the the geometric features of defects. Experimental results show that this new method can effectively suppress noise disturbance, improve the edge localization and detection accuracy, keep the edge information but simultaneously avoid the occurrence of false edges.