Research On Digital Image Copy-move Tampering Passive Forensics

In this age of intelligence,because of technological progress,mobile smart devices have been rapidly popularized,and digital images have become a popular information carrier.Digital media technology is also developing more and more rapidly.There are more and more digital image editing software on the market,with more powerful functions and simpler operations.It is becoming more and more difficult to rely on human eyes to judge whether a image is real,which has caused trouble to all sectors of society.As an effective means to protect the authenticity of digital image content,passive image forensics technology has been valued by many scholars.This thesis focuses on the common digital image tampering method-copy-move.Based on the current research results,two methods for detecting copy-move tampering are proposed.First,based on the Scale-invariant feature transform(SIFT)algorithm,some improvements were made in this thesis,and an adaptive threshold matching method was proposed.This method augments the image during the image preprocessing stage and selects appropriate parameters to facilitate subsequent SIFT feature extraction.Then extract SIFT feature points,use Generalized Tow Nearest Neighbor(g2NN)strict threshold matching for the first match,and use fuzzy c-means clustering to propose mismatched points,get the set of matching point pairs,then perform adaptive threshold matching to find more feature points,get another set of point pairs,and finally combine the two sets to get the tampered area.Experiments prove that the method has a good detection effect on various types of copy-move tampering oper-ations,such as geometric transformation or adding noise.Compared with SIFT algorithm,the performance has been improved.In view of the fact that the method in the previous chapter cannot meet the real-time require-ments,this thesis proposes another algorithm,Oriented FAST and Rotated BRIEF(ORB),where FAST stands for Features from Accelerated Segment Test and BRIEF stands for Bi-nary Robust Independent Elementary Features,feature extraction algorithm based on mul-tiscale Discrete Wavelet Transform(DWT).This method takes strengths of DWT and ORB,and achieve scale transformation by constructing a scale space.This method first performs multiple DWT decompositions on the image,then constructs a scale space,extracts ORB feature points and descriptors on the scale space,then performs g2NN matching,and finally uses Random sample consensus(RANSCA)to eliminate mismatches.The experimental re-sults are in line with expectations.For the existing copy-move tampering operation,the proposed algorithm can detect well and shorten the detection time compared to the ORB and the algorithm of the previous chapter.The algorithm can be used for scenarios that need to be detected quickly,which is consistent with current trend in the growth of technology.