Detection of median filtering based on ARMA model and pixel-pair histogram feature of difference image

Hang Gao, Tiegang Gao

Abstract: Median filtering is a widely used method for removing noise and smoothing regions of an image, and the detection of median filtering has drawn much attention from researchers of image forensic. A new robust detection scheme of median filtering based on pixel-pair histogram (PPH) and coefficients of autoregressive moving average model (ARMA) of difference image is proposed in this paper. In the proposed scheme, the PPH and ARMA are extracted from the difference image in four directions; the generated PPH-ARMA feature of 396 dimensions can effectively be used to detect the median filtering. In order to verify the effectiveness of the proposed scheme, a series of experiments on single database and compound databases are conducted, and the experimental results show that, the proposed scheme outperforms many existing algorithms. Moreover, the suggested approach achieves best performance in single dataset and multiple compound datasets compared with state-of-the-art methods, especially for strong JPEG compression and low resolution images.