The Blind Forensics Of The Copy-Paste And Splicing Image

Digital images as an important carrier of information, is widely used in all areas of daily life. However, with the development of modern science and technology, more and more image processing software lead that forgery of digital image is vnlnerable. So, the authenticity of the image information faces great challenge.Forensics of digital image is a novel technology, there are two methods in detecting image forensics: active forensics and blind forensics. A disadvantage with active forensics is that some prior information is embedded into an image for the authenticity of image, so the method limits the advance of researching. Blind forensics of digital image overcomes the limitations to a certain extent of active forensics. The Copy-Paste forgery and the image splicing operation are common forensics, it is particularly important to solve the problems in the passive blind detections of the Copy-Paste and Splicing.In order to conceal or enable a more realistic image, Forger forged a series of preprocessings which are resulting of low efficiency during the match resulting in process of the Copy-Move forgery. In addition, the searching of matching needs large amount Memory usage, high spatial complexity and a lot of time. In order to solve these problems, in this paper we describe an effective method to detect Copy-Move forgery in digital images. This method works by first extracting the Scale Invariant Feature Transform descriptors of an image and by seeking for approximate nearest neighbor based on product quantization.The method of approximate nearest neighbor search is to decompose the space into a Cartesian product of low dimensional subspaces and to quantize each subspace separately. The Euclidean distance between two vectors by computing the Asymmetric distance computation is used to determine the similarity of the region. Experimental results show that our approach can correctly detect the Copy-Move Forgeries which are preprocessed by different methods and decrease the memory usage and the complexity of learning the quantizer, at the same time, reduce the searching time.Splicing blind forensics of extrating the image features in the space domain is complicated and need a long time, in addition, the algorithm detection rate is not high. In order to solve these problems, in this paper we classify images by extracting space-frequency domain characteristics when using Empirical Mode Decomposition, and then obtain the optimal classifier parameters based on cross-validation method. Experimental results show that our approach reduce time and improve the accuracy of detection.