Research On Digital Image Forensics Based On Gamma Transformation

Image manipulation identification and tampering detection are two of the most important parts in digital image forensics.Image manipulation detection may contribute to the discovery of image operation history and the detection of image tampering,while image tampering detection is of great significance for maintaining social stability and safeguarding state security.This thesis focuses on the parameter estimation of gamma transformation and the detection of image splicing tampering based on gamma transformation identification.The mainly works of the thesis is summarized as follows:1.The property of gamma transformation and its effect on image histogram are analyzed.First of all,monotonicity and convexity of the basic gamma transformation are analyzed.Then by combining the properties of rounding operation and basic gamma transformation,the effect of gamma transformation on image histogram are descripted that there emerge new zero-value histogram bins and peaks after image gamma transformation,and the number and locations of zero-value histogram bins are closely related to gamma transformation parameter.2.A method for parameter estimation of image gamma transformation is proposed based on the relationship between the locations of zero-value histogram bins and gamma transformation parameter.Firstly,the relationship between the number of zero-value histogram bins in a selected histogram interval and the parameter is exploited to determine the parameter range.Then two candidate parameters are used to gamma transform a referring image,and the intersection of zero-value histogram bin locations sets of two transformed images is constructed.Finally,the parameter is estimated by matching the locations of zero-value histogram bins.The experimental results validate the effectiveness of the proposed method,which outperforms the previous works.The applications of image inverse gamma transformation show that the visual quality of gamma transformed image is restored by inverse gamma transformation with the estimated parameter3.A splicing tampering area location method is proposed based on the identification of gamma transformation.Firstly,the classifier is trained by a feature vector consisting of the number,the first position and the last position of zero-value histogram bins,the average distance between zero-value histogram bins and the left end of preprocessed histogram,and the weighted sum of characteristics function coefficients of adjacent histogram differences.Then,the image is divided into overlapping blocks based on sliding window,and the probability of each block belonging to a gamma transformed image is estimated by the trained classifier,based on which the probability of every pixel being gamma transformed is calculated.Finally,the splicing tampered area is located according to the pixel probabilities.The experimental results show the proposed method is superior to that of the previous methods.The sliding window based on image blocking contributes to the location of spliced area in pixel level instead of in image block size level.Meanwhile,the proposed splicing detecting method is robust to median filtering,a widely used anti-forensics tool,which is significant to image forensics.Finally,conclusions and discussions are made on the paper,and the fields for further research are prospected.