In view of error diffusion halftone image, an inverse halftoning algorithm based on adaptive median filter is proposed, which improved from the traditional low-pass filtering inverse halftoning algorithm. Firstly, the algorithm smoothes the halftone image A with a Gaussian low-pass filter whose size is 11×11 and variance is 1.3, obtains the image B by adaptive median filter and smoothes non-impulse noise while trying to preserve the details. Secondly, gets the edge of image B by band-pass filter and the edge image F by de-nosing and enhancing the edge of image B. Finally, obtains the final inverse halftoning image by integrating F and B in a certain gain multiple. The experimental results show that the algorithm can not only make the image smoothly but also retain the details of the image, and the PSNR and IN of the processed image is better than the traditional low-pass filtering.