In order to slove the low contrast and image blur in X-ray fluorography caused by noise and low illumination conditions, a new noise detection composed of region detection and degree detection is presented to distinguish between noisy pixels and noise-free pixels in the original image. Then the similarity degrees are calculated between each pixel and median value in the region of neighborhood and considered as the weighted value. Eventually, the noisy pixels are improved by the fuzzy weighted median filter while the noise-free pixels are not modified. The experiment results show that the proposed method can suppress the noise and preserve the edge and detail information.