Copy-move Forgery Detection Based On KAZE And Polar Harmonic Transformation

In the past few years,with the prevalence of image acquisition equipment and image editing software,users can easily forge digital images,which is subverting the traditional concept of "seeing is believing".Therefore,there is an urgent demand for searching effective techniques to identify the authenti city,integrity and aboriginality of digital images.Especially,among all kinds of image content tampering methods,image copy-move forgery is the most common one.It refers to copy a part of the image and paste it into another part of the image.Generall y,it is often used in combination with image blurring,rotation,scaling and other operations,so as to achieve the goal of not being discovered visually.Hence,the research of passive forensics for image copy-move is not only facing huge technical challenges,but also has strong practical application value.On the basis of the artificial and natural interference of existing tampering methods,two different forensic algorithms are proposed.The main work and innovation of this paper are summarized as follows:Aiming at the problem that some tampered images have obvious brightness changes and insufficient extraction point with the single texture,this paper proposes an efficient copy-move tamper detection algorithm based on equalization preprocessing and KAZE features.Firstly,equalization histogram is used to enhance the global contrast of the image and extract the feature information in detail.Next,KAZE based on non-linear spatial filtering is used to remove noise while retaining the edge details,which has a good effect on the problem of insufficient amount of feature points.Finally,a series of operations such as preliminary matching,feature points clustering,stereo matching and morphological operation are performed to accurately locate the tampered region.The experimental results show that the algorithm can not only detect single or multiple tampering of the image,but also precisely locate the forgery regions.At the same time,the robustness of the algorithm to different types of post-processing is verified,including brightness change,image blurring,JPEG compression and color distortion.A detection algorithm based on polar harmonic invariant moments and improved dense block matching is proposed to solve the problem of high false-positive rate in natural interference,i.e.images with similar but genuine objects(SGO).Firstly,the image is divided into overlapping blocks and computing the feature information of each block by polar complex exponential transformation.Secondly,the offset field distribution of the image is obtained by fast and efficient block matching algorithm.Eventually,the accidental matching is eliminated by dense linear fitting,and threshold interception and morphological operation are concatenated to generate the final mask.The experimental results show that the overall performance of the algorithm in this paper is better than that of other algorithms for data sets with SGO images and forged images,and it also has good detection results for other common data sets and var ious post-processing operations.Up to present,copy-move forgery still lacks the ability to post-processing attacks and the natural interference of duplicated images.Therefore,it is hoped that algorithms and experiment results present in this paper will promote the research of copy-move forgery detection in the future.