LivingKnowledge goal is to bring a new quality into search and knowledge management technology for more concise, complete and contextualised search results.
The paper “Two Decision Fusion Frameworks for Image Forensics” co-written by M. Fontani, A. Costanzo, M. Barni, T. Bianchi, A. De Rosa, A. Piva has been published in The National Telecommunications and Information Theory Group GTTI, in Messina and Taormina, Italy, on June 20-22, 2011.
Image forensics research has mainly focused on the detection of artifacts introduced by a single processing tool, thus resulting in the development of a large number of specialized detection algorithms. In tamper detection applications, however, the kind of artifacts the forensic analyst should look for is not known beforehand, hence making it necessary that several tools developed for different scenarios are applied. The problem, then, is to devise a sound strategy to fuse the information provided by the different tools. In this paper we introduce two theoretical frameworks, based on Dempster-Shafer’s Theory of Evidence and on Fuzzy Theory respectively, to perform the
fusion of heterogeneous, incomplete or conflicting outputs of forensic algorithms. Both models are easily expandable to an arbitrary number of tools, do not require tools output to be probabilistic and take into account available information about tools reliability. To validate the proposed approaches, we carried out some experiments addressing a simple yet realistic scenario in which three forensic tools exploit different artifacts introduced by double JPEG compression to detect cut&paste tampering within a specified region of an image. The results we obtained are encouraging, especially when compared with the performance of a simple decision method based on the binary OR operator.