Research On Passive Forensics Methods Of Digital Image Seam Carving Forgery

With the advent of digital era and the rapid development of the Mobile Internet,digital images play an increasingly important role in our lives.However,various simple and easy-to-use digital image editing software make it easy for ordinary users to tamper with the content and size of an image,and the tampering technique is not recognized by human eyes.The authenticity and integrity of digital images are being questioned,Malicious digital image tampering obstructs forensic appraisal,which seriously threatens social fairness and justice.Therefore,the research on digital image forensics has important practical value and practical significance.To deal with image tampering operation using seam carving,the change of local texture and the correlation between pixels before and after tampering are analyzed,and a forensic method based on LBP and Markov features is proposed.Firstly,the LBP operator is exploited to convert the image from pixel domain to LBP domain.After that,the 2D JPEG matrix is extracted by JPEG compression.Then its first-order difference matrices are calculated in the horizontal,vertical,diagonal and minor diagonal directions respectively.Finally,the Markov features are gained from the difference matrices in each direction and used to identify whether an image is suffered from seam carving by the SVM.In addition,considering that the Markov feature is only the difference of adjacent elements and can not well reflect the change of the image when the tampering ratio is large,this thesis adopts the extended Markov feature.The extended Markov feature is combined with LBP for detection.Different from the previous algorithm,the Markov feature and extended Markov feature are extracted respectively from the first-order differential matrix in four directions,and the fused features are classified and detected by SVM.The experiment result shows that the proposed algorithm is superior to the traditional Markov feature algorithm and other existing seam carving detection algorithms.Especially when tampering is small scaling seam carving,the detection efficient will obviously improved.In addition,the excellent detection rate on different databases reflects the adaptability of the algorithm in detecting tampering.