Digital Image Tampering Detection Research Based On Blind Forensics Technology

21st Century, with the continuous development of computer technology, the number of digital images is increasing very quickly. At the same time, image processing technology also achieves great progress. We can see many images which processed by image processing software like Photoshop everywhere in our lives. Some of them even involve some sensitive fields like politics, military, court, which make great trouble to people’s life. Due to the problem, digital image forensics comes out. With people attach importance to this field, more and more specialists throw themselves into the research of this field domestic and overseas. In recent years, digital image active forensics technology is grown, but because of the limitation, it can’t be promoted. Therefore, currently more specialists are researching digital image passive forensics technology which called blind identification technology.Blind identification technology is a focus in the study of this article. Around the theme, we introduce the research status of the blind identification technology at home and abroad and analyze the existing problems. Then we give the classification of this technology. Synthesis of tampering is the most common means of tampering in life. Depending on the source of the tampered area, it is divided into the same image synthesis of tampering and different images synthesis of tampering. The same image synthesis of tampering also is called copy-move tampering. After madding analysis and research to this tampering method, we put forward a more efficient algorithm.Previous detection algorithms can detect image tampering region, but their pervasive problem is the low efficiency because of large amount of calculation. Aiming at this problem, we put forward a copy-move tamper detection algorithm based on improved singular value decomposition. In order to reduce the number of image blocks, the algorithm extracts the low frequency component by using discrete wavelet transform(DWT) before the image is partitioned. At the same time, it improves these image block feature vectors which are obtained by singular value decomposition by giving up those values that are not important to reduce the dimension of feature vector. In addition, in the process of image blocks matching, we add the operation of discriminating similarity of vectors by Euclidean distance in order to improve the accuracy of the algorithm. Through many times experiment tests, it shows that the algorithm can effectively detect copy-move tampering area, and costs short time. So its detection efficiency is high. After adding a certain amount of noise in images or JPEG compression, the algorithm’s detecting result is still good. So the robustness of the algorithm is also good.