Font Size: a A A

Research On Image Tampering Forensics Based On Sensor Pattern Noise

Posted on:2018-11-29Degree:MasterType:Thesis
Country:ChinaCandidate:H L GuoFull Text:PDF
GTID:2348330536485992Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
With the wide popularity of digital cameras,digital camcorders and smart phones,people have entered an area of reading pictures.In order to meet people's requirements for image editing,a variety of image-processing software tools powerful and easy to operate have emerged.The appearance of these tools also have brought a lot of tampered photo images flooding in the Internet,news and other mass media,and inevitably trigger a serious crisis of confidence to the society.As digital photo images becoming an important information carrier,how to ensure their integrity and authenticity in the storage,transmission and dissemination has become a research hotspot in the field of digital image forensics.Photo image tampering forensics algorithms based on the sensor pattern noise(SPN)are effective to various tampering operations,such as copy-paste,splicing and blurring,so the algorithms have been widely concerned by researchers in recent years.In this paper,the mechanism,extraction methods of the SPN and the SPN-based tampering forensics algorithms are analyzed in depth.In summary,the research work is carried out as follows:Firstly,an adaptive threshold detection algorithm based on texture complexity is proposed,considering the existing image tampering detection algorithms based on the SPN often have high false alarm rate in strong texture areas,inaccurate tampering location and low detection efficiency.According to the Nyman-Pierson criteria,the correlation judgment thresholds corresponding to different texture complexity are determined,and then the fitting function of thresholds and texture complexity is set up.In the detection,the non-overlapping-block strategy is used to calculate the correlation between the SPN of the test image and the SPN of the inquiry camera.According to the texture complexity of the corresponding test image blocks,the appropriate threshold is selected to judge the correlation.As a result,the false alarm rate in the real areas of complex texture is reduced,and the area of the tamper is roughly determined.Then,the fast ZNCC algorithm is used to calculate the correlation of the corresponding pixels in the roughly localized regions between the SPN of camera and the SPN of the test image.Finally,the accurate localization is realized.The experiments on the mobile phone image database show that the detection rate of the proposed algorithm is 98.8%,while the false alarm rate is only 1.897%.Compared with the existing methods using a fixed threshold,the proposed algorithm using adaptive thresholds effectively reduces the false alarm rate in the complex texture region,and locates the tampered region accurately.At the same time,the efficiency of this proposed algorithm is increased by 26 times compared with the conventional sliding-window-based algorithms.Secondly,aiming at the unsatisfied result of tampering detection because of the low SPN quality of the test image,a method of the SPN-preprocessing based on non-linear guided filter is proposed to improve the accuracy.The SPN of the test image is susceptible to image content,camera CFA interpolation noise,JPEG compression noise and other random noises during the extraction process,and these noises will destroy the original characteristics of the pure SPN.The non-ideal SPN affects the quality of tamper forensics.Using the high-quality camera SPN to deal with the non-ideal SPN of test image based on non-linear guided filter proposed in this paper,the algorithm eliminates the interference noise effectively,get a high-quality SPN of test image and improve the accuracy of tamper detection.The experimental results show that the proposed algorithm can effectively improve the detection accuracy of tamper block with different sizes and the maximum increase reaches 4.41% compared with the existing algorithms.In addition,the proposed algorithm also improves the robustness against JPEG compression.
Keywords/Search Tags:Digital image forensics technology, sensor pattern noise, tamper detection and localization, adaptive threshold, non-linear guided filter
PDF Full Text Request
Related items