Use this URL to cite or link to this record in EThOS:
Title: Digital image forensics based on sensor pattern noise
Author: Lin, Xufeng
ISNI:       0000 0004 6348 5623
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Access from Institution:
With the advent of low-cost and high-quality digital imaging devices and the availability of user-friendly and powerful image-editing software, digital images can be easily manipulated without leaving obvious traces. The credibility of digital images is often challenged when they are presented as crucial evidence for news photography, scientific discovery, law enforcement, etc. In this context, digital image forensics emerges as an essential approach for ensuring the credibility of digital images. Sensor pattern noise mainly consists of the photo response non-uniformity noise arising primarily from the manufacturing imperfections and the inhomogeneity of silicon wafers during the manufacturing process. It has been proven to be an effective and robust device fingerprint that can be used for a variety of important digital image forensic tasks, such as source device identification, device linking, and image forgery detection. The objective of this thesis is to design effective and robust algorithms for better fulfilling the forensic tasks based on sensor pattern noise. We found that the non-unique periodic artifacts, typically shared amongst cameras subjected to the same or similar in-camera processing procedures, often give rise to false positives. These periodic artifacts manifest themselves as salient peaks in the magnitude spectrum of reference sensor pattern noise. We propose a spectrum equalization algorithm to detect and suppress the salient peaks in the magnitude spectrum of reference sensor pattern noise, aiming to improve the accuracy and reliability of source camera identification based on sensor pattern noise. We also propose a framework for large-scale image clustering based on device fingerprints (sensor pattern noises). The proposed clustering framework deals with large-scale and high-dimensional device fingerprint databases and is capable of overcoming the NC >> SC problem, i.e., the number of cameras is much higher than the average number of images acquired by each camera. Additionally, for the task of image forgery detection based on sensor pattern noise, we propose a refining algorithm to solve the missing detection problem along the boundary area between the forged and non-forged regions. The proposed algorithms are evaluated on either a public benchmarking database or our own image databases. Experimental results, as well as the comparisons with state-of-the-art algorithms, confirm their effectiveness and robustness.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: TA Engineering (General). Civil engineering (General)