Blind Forensic Of Digital Images Based On The Optical Properties Of Images

As the development of image processing technology and the popularization of image processing software, people can tamper images easily and the tampered images will not leave any traces of tamper. Some malicious image attacker use forged images for military or political purpose, which has brought bad influence and caused a crisis of confidence. Therefore, the detection of the authenticity of digital images has become a hot topic in academic research in recent years. This paper designs a novel blind forensics method of digital images based on camera characteristics consistency. In this paper, we exploit imperfection in camera imaging process and consider the radial distortion parameters as a special natural watermark.Firstly, this paper explains the necessity of research on blind forensics of digital images, illustrates the main problems and classification of current digital image forensics technology. Moreover, this paper introduces some main research teams in the field of digital image forensics and gives a brief presentation on related research achievements.Secondly, we summarize the existing image forensics technology. Digital image forensics can be divided into initiative forensics and passive forensics, we describe them respectively. In initiative forensics technology part, we mainly introduce the principle and characteristics of digital watermarking technology and digital signature. In passive blind forensics part, this paper elaborates the problem needs to solve by blind forensics and proposes a universal framework of blind forensics research, which is the focus of the study in this article. Because the radial distortion emphasized in this article is a kind of image feature which is caused by the imaging devices, so the blind forensics methods based on camera characteristics consistency are expounded.Then, this paper establishes a mathematical model of camera imaging process and discusses the principle, classification and mathematical expressions of lens distortion based on this model. Combined with the geometrical optics principle and the mathematical model of the process of imaging, we deduce the polynomial model of radial distortion and archive the relationship between the physical meaning and mathematical expressions of radial distortion parameters. This paper also designs a set of detection methods to calculate the radial distortion parameters and discusses every step in detail and methods of implementation. The methods use improved non-maxima suppression of the magnitude of the gradient of image intensity in the direction of this gradient (NMS) with sub-pixel accuracy for edge detection, find the straight lines part of the images by Hough transform and extract the distorted points on them by defining a tolerance region_According to the coordinates of these distorted points we can archive the radial distortion parameters using iterative optimization method. At last, we can identify whether the images are tampered by judging whether the law of the distortion parameters in natural images is disturbed.Finally, the simulation of the method proposed in this paper is implemented to three main tampering means, which is images copy-paste, images cropping and computer generated image. The simulating results are analyzed. We summarize our work in this paper and propose the development direction of our research in the future.