Forensics of Operation History Including Image Blurring and Noise Addition based on Joint Features

Abstract

Multi-manipulation is becoming the new normal in image tampering while the forensic researches still focus on the detection of single specific operation. With uncertain influences laid by pre-processing and post-processing operations, the multi-manipulation cases are hardly identified by existing single-manipulated detection methods. Here come the studies on image processing history. In this article, a novel algorithm for detecting image manipulation history of blurring and noise addition is proposed. The algorithm is based on the change of attributes correlation between adjacent pixels due to blur-ring or noise addition. Two sets of features are extracted from spatial domain and non-subsampled contourlet transform (NSCT) domain respectively. Spatial features describe the statistical distribution of differences among pixels in neighbourhood, while NSCT features capture the consistency between directional components of adjacent pixels in NSCT domain. With the proposed features, we are able to detect the particular processing history through supporting vector machine (SVM). Experiment results show that the performance of proposed algorithm is satisfying.