Research On Intelligent Forensic Technology Based On Computer Image

With the development of information technology, Digital image has been one of most important media for information transmission. And detecting the digital image for true or false has been the hot research area. Today, there're two directions in the research of detecting the digital image: passive digital image forensic technology based on digital water print and active digital image forensic technology. The passive technology is about to adding the known information into the digital image to accomplishing the forensic aim. And The active technology is to detect the unknown source digital image, the main research direction has been two divisions, one is aim to the natural image modified by computer software for forensic detecting, the other is aim to the digital image made by computer software.In this paper, a general digital image forensic detecting method is proposed, in which a multi-dimension characteristic vector of image is used to form a 2-class classified model. The characteristic vector of image is composed of image space feature,and a set of characteristic features based on higher-order statistics. Besides, a improved algorithm is also proposed, in which characteristic function and higher-order statistics are combined to form a new characteristic vector of image.Both of the forensic detecting methods focus on acquiring effective characteristic features. First, the characteristic features of training image are acquired to form training samples, which is used to train a classifier(AdaBoost) in order to form an effective 2-class classified model. Then, the same characteristic features of image to be tested are input to the classified model, which can tell whether a image is computer image. Experiment show that both of the schemes work well, they can detect the computer image efficiently.