Abstract: This paper mainly studies the edge detection of noisy images. An edge detector based on the fuzzy theory is presented. The algorithm first makes the pixels in the window as a fuzzy set, and divides two sub-sets by their membership functions, and then iterates in the two sub-sets, lastly detects the edge by the difference of the two values from the two sub-sets. The computer simulation shows that the edge detector is good for attenuation not only of Gaussian noise but also of impulse noise and mixture noise.