An adaptive median filtering method based on SNR estimation of single image is presented in this paper. The proposed method first innovatively estimates SNR of an image degraded by salt & pepper noise, and then adaptively chooses window size of median filter. Estimation of single image's SNR provides guidance for the judgment on noisy or non-noisy image, and even noise density. This is helpful for dealing with salt & pepper noise, such as adaptive determination of window size in median filter. The new method was compared with classic median filter with fixed window size on a variety of images, and experimental results show its effectiveness.