Digital Image Forensics with Statistical Analysis
Xunyu Pan · 2015
A large number of forensic methods have been developed in the past decade to answer a broad range of forensic questions. Most image forensic tools can be divided into only two simple categories: semantics-based detection and non-semantics-based detection. This chapter first focuses on the non-semantics-based detection techniques as majority of existing image forensic tools fall into this category. The non-semantics-based detection tools mostly rely on the modelling of statistical patterns of the image using signal-level information. The chapter introduces several recently developed techniques to address two critical topics in the field of multimedia security: detecting region duplication and exposing splicing forgery. It talks about a method for reliable detection of duplicated image regions and an effective image splicing detection algorithm. More realistic case studies for these two techniques are demonstrated. These techniques are further extended to expose forgeries in audio and video signal.