A Survey Of Passive-blind Forensics Technology About Artificial Blur In Digital Image Tamperings

With the widespread use of image editing software, it is increasingly easy formodifying digital images, whose authenticity is facing a threat. If tampered imagesappear in some areas where the primitiveness of image is strongly required, it maylead to a negative social influence, so it is important and significant to the digitalimage forensics technology for identifying the authenticity of image. Therefore，wehave a deep research on the passive digital image forensics technology in this paper.Digital image passive forensics is a detection technology for the tampered images.The blur operation is the most common way of polish after images are tampered. Nowthe research of the detection of artificial blur tampering in digital images has achievedsome results. Some scholars have proposed different detection algorithms fromdifferent views such as the image space domain, frequency domain, edge features andother aspects. The existed algorithms can locate the fuzzy regions, at the same time,there still exists some problems and defects. In this paper, we propose three artificialblur detection algorithms based on full consideration and analysis of the existedalgorithms. The main work of this thesis is as follows:1. Briefly introduce the current common methods on digital image tampering, anddetailedly analyze the current research of digital image forensics technologies.2. A detection algorithm of artificial blur based on gray-scale anomaly isresearched. By observing and analyzing the impact of blur operations to gray scale ofthe image in the space domain,we summarize the regularity and achieve blur detection.The method is simple and having good effect in detecting.3. A blur detection algorithm based on blur estimation and improved abnormaltone is researched. The algorithm is an improved version of an existing algorithm.Experimental results show that it significantly improves the accurate detection rateand expands detection range of images.4. An artificial blur detection algorithm based on PRNU is researched. PRNU is aspecific fingerprint of a digital camera. This detection model is based on the impact offuzzy operation on PRNU. Campared to the existing algorithms, it can effectivelyavoid the impact of image content on the detection, moreover, it can achieve a higheraccurate detection rate.In this thesis, we propose three blur detection algorithms that can effectively locate the artificial blur existing in the digital images and also have some degrees ofrobustness, which they can withstand some of post-processing operations, such asJPEG lossy compression with different quality factors, image rotation with differentangles.