Research On Blind Image Forensics For Cloning And Splicing

Today,the modification of digital images have become more and more easier and reality.The powerful image processing tools not only bring high efficiency to the people in living and working,but also make the image easier to tampering and forgery.Some fake digital images have reached a believable level which bring great affect to the our society and disrupt social order.The digital image forensics technologies have become a crucial and hot issue in the field of the criminal investigation,security,publicity,intellectual property protection.In this thesis,we analyzed the existing image forensic methods,and focus on the image forensics of cloning and splicing from the aspects of theoretical innovation,algorithm improvement.Moreover,we have made a number of improvement to some existing schemes.The main research contents are list below:(1)We have proposed an automatic and efficient forensic method for copy-move forgery detection based on DWT-FWHT.It can work in case of complete absence of digital watermarks or signatures.To make the method more efficient compared with previous works,the proposed method uses DWT to reduce the image size,and extract the feature in the overlapping blocks by FWHT instead of the DCT.To speed up the method,MHJ algorithm is used in the matching process.The experimental results show that the proposed method can accurately and quickly detect the duplicated regions.And the processing time reduced significantly as against other classic methods,while the accuracy still remains a high level.The method works well in spite of the presence of blur,noise and JPEG compression in a certain degree.(2)We have proposed a novel methodology to support image copy-move forgeries based on SURF technique.Although block-based approaches seem effective to detect duplicated regions,they are incompetent to deal with the forged image which may manipulated by multiple kinds of transformations such as scaling,rotation,flipping,blurring,illumination change and multiple cloning.Therefore,we make use of SURF feathers.We find the keypoints and extract the features in the image at first.Second,we match the keypoints based on kd-trees algorithm with the contrast of feature.Third,we clustered the keypoints by using the agglomerative hierarchical clustering algorithm.Finally,we estimate the parameters of affine transformation between matched keypoints by feature-based method.(3)Splicing forgery is a common type of forgery where forged region is copied and pasted from another image,and is hard to be detected due to the absence of reference object.Furthermore,the forged regions are always subjected to various image transformations in order to conceal the tampering.We have described a novel forensic method to exploit splicing forgery.The detection method is performed on the basic principle that a natural penumbra would contain some properties which the forged one is lacking.we firstly extract the shadow boundaries and the penumbra shadow regions in an image.Then,a simple and efficient method based on the Growth Rate of the Penumbra Width(GRPW)is used to exposing the splicing forgery.(4)We have proposed an improved method for splicing image forensics based on shadow model.The detection method is performed on the basic principle that a natural shadow should contain some properties which the forged one is lacking.First,we analyze a variety of classic shadow models,and establish a shadow model which conformed to be more suitable for image forensics.Second,we make use of color invariant features to extract and locate shadow automatically.Third,we estimate the shadow scale factors by using the energy minimization method and removing the interference points.Finally,the fake shadow can be determined by detecting the abnormal value.The experiment results demonstrate the efficiency of the proposed method.