The Research On The Forgery Detection Of Composite Digital Image

The image tampering technology developed rapidly caused by the popularity of digital cameras and the widespread use of image editing software. The initiative digital image forensics technology by using digital watermarking and digital signature which needs to embed information in the image in advance has shown some limitation of application in the new environment. Therefore, the passive digital image forensics technology, known as passive blind forensics technology, has become a research hotspot in recent years. The digital image blind forensics technology which was studied begin with researching tampering methods conducts the forensics of the authenticity, integrity and original of digital image by analysing the image statistical features and the property of imaging equipment. Currently, the researchers both domestic and overseas concentrate their attention mainly on the forensic of the authenticity of digital image content and some effective detection methods were proposed which aimed at copy-paste tampering, re-sampling, JPEG re-compression and fuzzy retouching operation to the digital image.The composite tampering of digital image was the most typical and most common used type of tampering techniques, which includes the copy-paste operation in one picture and the splicing operation between different images. Moreover, most of the forged images on the internet experienced such tampering. So, two methods for detecting such image tampering operation was proposed in this paper as follows:Detecting copy-paste tampering in one picture based on truncated singular value decomposition(TSVD) and K-D tree. Perform discrete wavelet transform on the image to get the low-frequency component firstly, and choose the truncated singular value of sub-block as the feature of the image which expresses image feature unique and in low-value dimension. Fast and effective searching algorithm using K-D tree will take place of conventional lexicographic sort method. With the cooperation of Euclidean distance used as similarity measure and Chebyshev distance used as measurement of offset of coordinates, the copy and paste regions can be detected accurately and speedy.On the basis of the research on features of Human Visual System and JPEG blocking artifacts, a measurement of blocking artifacts based on the visibility function of Human Visual System (HVS) is proposed in this paper. By using the inconsistence of blocking artifacts between tampered region and background region, detection can be achieved on image. The mean standard variance of Brightness slope of adjacent image blocks is used as the measurement of blocking artifacts. Meanwhile, the mask to blocking artifacts of the background activity and brightness is also taken into consideration. Then applied it to detect composite JPEG image, the validity of the algorithm was shown through experiment and the results was easy to identify.The detection principle and detailed steps of each algorithm mentioned above are discussed in detail in this paper. A series of experiments have shown the effectiveness and stability of the algorithm in actual detection of tampering images. A forecast to the development of the digital image forensics technology was made finally.