Research On The Detection Of Digital Image Tampering Based On Copy And Paste

At present, because of the large amount of information, vivid characteristics, thedigital image as a good carrier of information transmission, has been widely used inpeople’s daily life and work, which makes people’s life more colorful. However, thesophisticated digital image editing software, such as Photoshop, ACDsee, makes the imagemodification operations to be easy. Ordinary computer users can easily modify the contentof a digital image so that they can mix the spurious with the genuine. In recent years,frequent exposure of fraud has served as the warning to us. If some malicious tamperingimages are abused, it will have a bad effect on personal privacy, the social stability, andeven may have adverse impact on the nation’s politics, military etc. Therefore, with thedevelopment of today’s society, it is the time for digital image authenticity identification todistinguish between true and false photos. Image forensics technology has become moreand more important, which has become a very hot research direction.Digital image tampering has a wide variety of ways, including copy paste is one of themost important means of tampering with digital image tampering. According to thedifferent sources of replication region, digital image copy paste can be divided into thesame image copy paste tampering and ISO image copy paste tampering. The detectionmethod of the digital image tampering mainly has: the exhaustive search method, the self-correlation detection method, based on the DCT method and based on the PCA method andso on. These methods detect tampering of digital images from different aspects, but thereare many problems in the aspects of the high detection complexity and no high accuracy.This paper mainly studies the copy paste forgery detection method in the same image.And based on the study of detection methods, it puts forward a copy paste tamperingdetection algorithm based on PCA clustering. The algorithm is to separate the tamperedimage firstly, and each sub block image is a sample data. Then all the sub block images are formed the analysis sample collection. Secondly, the sample collection is analyzed on theprincipal component. The sub block image set maps to the main feature space, andanalyzes clustering of image collection in the main feature space. Finally, the image ineach cluster similarity is detected by similarity judgment whether the image is the copypaste forgery. The experimental results show that the algorithm can effectively detect thesame image in the copy paste forgery.