Blind forensics tool of falsification for RAW images
- Resource Type
- Conference
- Authors
- Doan, Thi Ngoc Canh; Retraint, Florent; Zitzmann, Cathel
- Source
- 2017 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) Signal Processing and Information Technology (ISSPIT), 2017 IEEE International Symposium on. :018-023 Dec, 2017
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
Robotics and Control Systems
Signal Processing and Analysis
Forensics
Digital images
Cameras
Tools
Numerical models
Forgery
Detectors
Digital forensics
blind image forgery detection
hypothesis testing
image models
- Language
This paper presents a novel method for blind forgery detection of natural image in RAW format. The approach is based on a statistical noise model of natural RAW images. This model is characterized by two parameters which are used as a fingerprint to falsification identification. The identification is cast in the framework of the hypothesis testing theory. For practice use, the Generalized Likelihood Ratio Test (GLRT) is presented and its performance is theoretically established in case of unknown parameters where an estimation of those parameters is designed. Experiments with simulated and real images highlight the relevance of the proposed approach.