Research On The Video Tampering Detection Methods Based On The Inter-frame Correlation

Nowadays,with the improvement of the ability of digital information processing,people can get digital information in a very simple way.Due to the widespread use of smart phones,cameras and other digital devices,more and more pictures and videos are produced and edited by ordinary users,and are uploaded to the Internet for sharing.The usability of video editing and modifying technology has led to the problem of video copyright infringement and so on.It has become a challenging task to identify the original video from the illegal version.The main content of this paper is to detect the inter-frame forgery and detect the intra-frame forgery in digital videos,and the main results are as follows:A series of video inter-frame tampering detection method based on global texture features of video frames are proposed.In the proposed method,the low frequency component of the video frame series is extracted using the one-dimensional Haar wavelet decomposition,and the extracted low frequency series are used as tested video series.Then we extract global texture features by using GIST feature descriptor,and establish the detection algorithms for video frame copy,frame insertion and frame deletion tampering,respectively using Euclidean distance,the correlation coefficient and the local outlier factor.We define discrimination rules to realize video frame copy detection,frame insertion detection and frame deletion detection.This method is characterized by the use of one kind of feature to achieve three different interframe tampering detection tasks.The experimental results show that the proposed scheme has better performance in detection effect and robustness,and can accurately determine the video frames numbers to be copied and inserted into the video series.Considering a similar texture features between the original moving object and its copied version in video,we propose an intra-frame Copy-Move tampering detection and tampering localization method based on the texture similarity.In the proposed method,we still use the low-frequency component of wavelet decomposition as the tested video sequence.First,thek-means algorithm is used to cluster the GIST features for video frames,and video frames are divided into move object frames and others according to the clustering result.For the frame sequence that include move objects,the SIFT feature extraction algorithm is used to extract feature for each frame.According to the matching result,intra-frame Copy-Move tampering detection and tampering location can be achieved.The experimental results show that the proposed method has a satisfactory effect in detection accuracy.A method based on noise consistency for video tampering detection is proposed.This method can implement video frame insertion tampering detection and intra-frame tampering detection.First,the test video is divided into frames and converted from the RGB color space to the HSV color space to obtain the H channel frame sequence,Then by using Wiener filter to denoise for this frame sequence,frame noise residual sequence are obtained.If the frame insertion tampering is detected,the Spearman correlation coefficients of adjacent frame in noise residuals sequence are calculated directly.The average value is calculated,and the ratio of the average value to the minor value is compared with the threshold to determine whether there is any frame insertion tampering or not.If the intra-frame tampering is detected,the noise residual sequence should be processed in block and then detected by the same technique.This method has a good detection ability.