Research On Video Forgery Passive Authentication Technology

With the development of software and hard ware technology, network technology and business model, digital image techno logy has been greatly developed and video has become an indispensable part of social life. Video not only as an entertainment tool to enrich people’s lives, but also is one of the indispensable security and judicial tools of modern society. In solving judicial disputes, video can be used as one of the important means of proof. In fact, s urveillance video has become one of important source of judicial forensics. At the same time digital image processing technology and the corresponding software continue to emerge, videos can be scarcely distinguished from natural ones after the processing of image software, and there are no traces left in visual. For the simplicity of the tampered vid eo, many criminals may make the tampered video to spread rumors and blame others and other criminal acts. Therefore, video forgery detection has become one of the important directions in video research.By analysis and study on the video tampering passive detection results, this thesis has done the following work: Firstly, the thesis introduces the definition, classification and research status of video tamper passive detection technology in detail; Secondly, the thesis proposes an inter-frame tampering passive detection algorithm based on histogram intersection and an intra- frame tampering passive detection algorithm based on the improved three frame difference method.Inter- frame tampering passive detection algorithm based on histogram intersection: Inter- frame tampering techniques have two common characteristics, 1, the frames around the tampering have no continuity; 2, the frames around the tampering are usually with static background and no moving target. By using the characteristics of the frames around tampering, this thesis proposes an inter- frame tampering passive detection algorithm based on histogram intersection. In order to detect continuity of frames around the tampering, this thesis does an effective conversion for calculating the continuity between pixel belts, which greatly reduces the computing data. Because the frames around the tampering are usually with static background, this thesis adopts the histogram intersection method calculating the correlation between pixel belts. Therefore, the advantage of inter- frame tampering passive detection algorithm proposed in this thesis is that the algorithm complexity is low. The specific detection step is as follows: first, a pixel line will be obtained by extracting a line of pixels on the horizon(on the vertical) from every frame of a video sequence. Every four continuous pixel lines make up a pixel belt. Then, the histogram intersection will be used to calculate the correlation coefficients between pixel belts. If the video has been tampered with, the continuity of pixel belts correlation will be destroyed. This thesis tests videos with frame deletion, frames duplication and frame insertion. Experimental results show that the algorithm proposed in this thesis can recognize the tampering techniques, ensure the detection efficiency and effectively reduce the complexity of the algorithm.Video frame tampering detection algorithm based on improved three frame difference method: This thesis mainly studies the intra-frame moving target removal tampering detection. This detection has two steps: First, locating the suspicious tampering frames in the time domain, this thesis improved the frequency domain energy suspicious degree algorithm, which makes it more sensitive to the energy change of video image. Secondly, locating the tampering position in the airspace, this thesis proposes the improved three frame difference algorithm to locate the tampered position. The main detection process is: Firstly, by using the DC T, the video image is transformed to frequency domain. if the video has been tampered with, then sequence pictures of energy suspicious degree of sequential consistency will be destroyed, so the suspicious tampering frames can be detected. Then the improved three frame difference method will be used to extract of the residual moving foreground of suspicious frames, thus the tampering poison can be detected. Experiments show: In time domain, the improved frequency domain energy suspicious degree algorithm can locates the suspicious frames of video with different compression format; in the airspace, the improved three frame difference algorithm can locate the tampering with the specific position in frames.Finally, this thesis analyzes the advantages, disadvantages and the effectiveness of the algorithm. The factors and parameters influencing the results of experiments were also analyzed. At the end of the article, this thesis makes a prospect and summary on the video tamper detection method...