Font Size: a A A

Double Compression Detection And Tampering Localization Of JPEG Images

Posted on:2012-09-19Degree:MasterType:Thesis
Country:ChinaCandidate:L S DongFull Text:PDF
GTID:2218330368987803Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
With the popularity of digital cameras, digital image has become one of the most important media forms in modern communication because of its convenience, instant and easy to transport. At the same time, with the development and widely spreads of the image processing software, tampering a digital image becomes an easy job for non-specialist and the tampered image is hardly detected by the naked eyes. In recent years, more and more tampered images emerged in the field of political, judicial, news, etc. both in domestic and overseas, which has caused skeptical about the integrity and authenticity of the images. Thus, image forensics, especially those technics using only the image characteristics, has become a hotspot of study in recent years. In this paper, considering double compression effects on the statistical characteristics of JPEG images, we proposed two different double compression detection methods, and a JPEG image tampering localization model using double compression detection features. The main contribution of this thesis includes:(1) Proposed a double compression detection method based on the Gray Level Co-occurrence Matrix (GLCM) of the DCT (Discrete Cosine Transform) coefficients. Through deep analysis about the double compression effects on the DCT coefficients of JPEG images and those methods proposed to detect double compression previously, we find that some correlations between adjacent coefficients, which have been ignored in the previous works, can be used to increase the detection accuracy. Based on this analysis, we find that the Gray Level Co-occurrence Matrix, which is a second order statistics, can reflect the correlations and thus enlarge the double compression effects on DCT coefficients. Experiments show that this second order statistics do increase the double compression detection accuracy compared to the first order statistics.(2) Proposed a double compression detection method based on the Markov model of the first digits of DCT coefficients. After a comprehensive analysis about the double compression effects on the distribution of the first digits of DCT coefficients, we proposed to model the distribution of the mode based first digits of DCT coefficients using Markov transition probability matrix and utilize its stationary distribution as features for double compression detection. Experiments show the effectiveness of the proposed method, especially when the second compression factor is much lower than that of the first one, the detection results have a significant improvement. (3) Proposed a JPEG image tampering localization Model based on double compression detection. Most of the tampered image will be resaved in JPEG format to save the storage space while maintaining the image quality, which may leave double compression artifacts. Study on the tampered JPEG image shows that it can be divided into two parts:one is the original part that has double compression artifacts and the other is the tampered part without double compression artifacts, thus by detecting whether the image blocks have double compression artifacts, the tampered region can be determined. Experiment results show that the proposed model is effective under post-processing operations such as rotation, resizing, feathering, etc. What's more, the model gives promising results even when the tampered image has been compressed at a relatively low quality factor.
Keywords/Search Tags:Double Compression Detection, Gray Level Co-occurrence Matrix, Markov Model, Image Tampering Localization
PDF Full Text Request
Related items