A Content-Adaptive Median Filtering Detection Using Markov Transition Probability Matrix of Pixel Intensity Residuals

Abstract

Excessive dependence on digital images has raised the need for its forensic analysis. Median filtering, due to its nonlinear nature, is extensively used in image antiforensics to hide the evidence of image forgery. There exist several methods based on the Markov transition probability matrix to detect median filtering. In this article, to detect median filtering, a content-adaptive thresholding approach is applied on pixel intensity residuals and features are extracted using a Markov transition probability matrix. This modified approach provides better results even on low-resolution and highly compressed images in comparison to existing methods.