Blind Median Filtering Detection Using Statistics in Difference Domain

Recently, the median filtering (MF) detector as a forensic tool for the recovery of images' processing history has attracted wide interest. In this paper, we focus on two topics: 1) an analysis of the statistics in the difference domain of median filtered images; 2) a new approach based on the statistical characterization in difference domain to overcome the shortages of the prior related works. Specifically, we derive the cumulative distribution function (CDF) of first order differences based on simplifying assumptions, and also study the behavior of adjacent difference pairs in the difference domain for original non-filtered images, median filtered images and average filtered images. We then present a new MF detection scheme based on the statistics in the difference domain of images. Extensive simulations are carried out, which demonstrates that the proposed MF detection scheme is effective and reliable for both uncompressed and JPEG post-compressed images, even in the case of low resolution and strong JPEG compression.