| With the popularity of high performance personal computer and the emergence of advanced digital video editing software, and constantly upgrade, to modify a video is becoming more and more easy. Skilled, can even do it convincing results. If this technology is mastered and used in illegal activities, will have unpredictable consequences for society. Therefore, how to identify the primitiveness of the period of digital video has become one of the research focus of scholars. This article in view of the way of insert frame tampering with, respectively, from the angle of pattern noise and light intensity analysis be tamper with the video data characteristics, and applied to tamper detection for a video to be detected.Pattern noise is due to the image sensor (CCD) of non-ideal of interference and the introduction of additional data. It has distinct uniqueness. Based on the uniqueness of pattern noise, combined with the proposed a semi-supervised learning algorithm based on inverse gravity density, clustering analysis to measure the noise of the video, video detection is affected by different source frame tampering. First of all, to extract video three color separation after each frame of three color separation of noise, and then on the extraction of the noise mean and variance, with a six-dimensional vectors to represent each frame. Finally put forward by the use of a semi-supervised learning algorithm based on inverse weight density of data aggregation class whether video experience by the clustering results tampered with.Light intensity is an important information of video scene. Firstly, obtained through the data conversioned by the gradation of the video frame with the HSV color space, brightness and saturation light luminance characteristics of these features is then fused into a representative feature. Finally, analysis of each frame of light on behalf of the characteristics and the degree of deviation of the adaptive threshold detection results. |
PDF Full Text Request