Blind Authentication Of Copy-move And Blur Forgery In Digital Image

Image data as an important message carrier, has been widely used in news reports, intelligence information gathering, criminal investigation, security monitoring, medical and scientific researching. However, the advent of image processing software which are low-cost, high performance and friendly user interface lead that digital image is vulnerable to a malicious forgery. So the digital image authentication the current study has become a hot spot.Today, there're two directions in the research of detecting the digital image: active technology based on digital watermarking and passive-blind technology. The former works by inserting prior information into an image for the authenticity and integrity of image. Unlikely, the later approach works directly on the information of the digital image without any prior embedding information such as watermarking. Although passive-blind technology detecting digital image forensic is less developed than active technology, but it is bound to the leading detecting digital image forensic technology.First, in the report, we analyzed the changes of the image characteristics introduced by image captured and forgery theories, principles and effects of algorithms. The focuses of this report are detecting copy-move forgery and blur operation which are the most common operations in digital image forgery.The existing algorithms on copy-move detecting are quite computationally expensive for lexicographically order. In order to overcome the drawback, an autocorrelation detection methods based on the discrete wavelet transform is proposed in this paper. In this method, low-frequency sub-band wavelet decomposition which is best approximation of the image was extracted as a feature vector. So the method greatly reduced the dimension of feature vector and accordingly increased detection efficiency of the matching block.The existing splicing detection algorithms are unable to detect the polished digital image. In this paper, a detection method for blurring operation is proposed. By analyzing the image edge features and the characteristics of artificial blurring operations, homomorphic filter was used to enhance blur edge and contrast pixels between inside and outside. The approach is fast and accurate to detect and locate the forgery region.