Research On Inter-Frame Forgery Identification Based On Video Feature QoMSSIM

In recent years, with the development of network and multimedia technology, digital videos are widely used in news, finance, business and even as evidence in court. However, functional diversity and easy operation of the digital video editing software make it easier to tamper video. If malicious tampering video was captured and used by criminals, it may bring immeasurable negative impact on the whole society. Frame deletion and insertion of digital video is one of the most common tampering operations. Therefore, video inter-frame forgery identification has become a hot topic in video forensics.In this paper, some kinds of classic video forensics techniques for detecting inter-frame tampering are studied. On this basis, a new algorithm based on video inter-frame feature QoMSSIM is proposed to detect video forgeries attacked by frame deletion and insertion. The specific work of this paper is as follows:A new algorithm based on video feature QoMSSIM is proposed for inter-frame tampering detection. In this paper, MSSIM is used to measure the correlation between adjacent frames of video. MSSIM of adjacent frames is divided by another one, and then processed by normalization, quantization and dimensionality reduction. Finally, we get QoMSSIM as classification feature for video tampering detection. We put the majority of QoMSSIM into SVM (support vector machine) to train sample, and the remaining part of QoMSSIM is used to test classification accuracy of the proposed algorithm.In this paper, a large number of experiments are conducted to verify video tampering detection algorithm. Experimental results show that the proposed algorithm has a high detection rate in the classification of original and inter-frame tampering video, which is better than the algorithm based on optical flow feature. In terms of computational complexity, the proposed algorithm is also significantly lower than video tampering detection algorithm based optical flow feature. In the paper, the proposed algorithm can further classify frame insertion and deletion tampering video, and the detection accuracy is more than 92%.