Digital Image Blind Forensics Technology Research And Application

With the development of science and technology, the digital image is becoming more and more widespread in our daily life. Meanwhile, most of the ordinary digital cameras are not equipped with digital watermark or signature capabilities. The images in digital format can easily be tampered with, when human intervention or technical barriers are involved. If you do not keep a copy of the image for reference, it is difficult to verify its authenticity. So the proactive detection techniques are no longer useful for the authenticity verification of digital images. In the case that the image does not carry its digital watermark, the blind forensics, as a part of digital image forensics, play a vital role in the authenticity verification. The main work done is as follows:(1) For the blind detection of digital image copy-paste tampering, based on the analysis of SIFT algorithm and SURF algorithm, we preprocess the RGB space image with Vector-Order color edge detection operator, which can get image edge features. And combined with SIFT algorithm, which remains invariant under translation, scaling and rotation operations and maintains a certain degree of noise stability, we obtain feature sets and match feature points for tampered images. Finally we propose the SIFT-based color edge detection (CV-SIFT) algorithm. In this paper, we verify the efficiency and performance of CV-SIFT algorithm in a sample library of digital images. The feasibility and accuracy of the algorithm are testified by the experiment result.(2) For the detecting algorithm of computer-generated graphic, which meets the needs of digital image forensic, we analyze the computer image detection algorithm based on the fractal dimension, image noise and HSV models characteristic function, and improve this algorithm referring to the concept of classifier. The accuracy is testified in a sample library of digital images.(3) After analyzed the current demand for digital image forensics software, we develop the digital image forensics system on OpenCV and Visual studio6, and implement the algorithm of (1) and (2), in the blind detection modules for natural image and for computer-generated graphic. We build a digital image blind forensics system with high efficiency. In this paper, we verify the detection accuracy and practicability of the system by the experiment, and apply the system to the processing of digital image blind forensic.The algorithm in this paper improves the efficiency of copy-paste detection of color digital images. Meanwhile, we propose an efficient system to meet the current needs of the digital image forensics industry, which achieves a rapid detection for natural image with copy-paste tampering and CG image.