基于MSSIM商一致性的视频插帧和删帧篡改检测

插帧和删帧是常见的视频帧间篡改方式，针对此研究问题，提出一种基于结构相似度均值( MSSIM)商的一致性检测算法。对于一段连续拍摄的视频，由于视频内容的连续性，相邻帧之间的MSSIM商具有连续现象。而对于经过插帧或删帧篡改的视频，其篡改点位置的MSSIM商会发生突变，从而破坏其一致性。为此，该算法首先计算每相邻两帧间的MSSIM值，之后对相邻MSSIM求商并以此作为特征，而后两次利用切比雪夫不等式及阈值法对提取特征进行异常点检测，从而实现对视频插帧和删帧的篡改检测及定位。实验结果表明，该算法对视频插帧和删帧篡改具有较高的检测率。

Video-Frame Insertion and Deletion Detection Based on Consistency of Quotients of MSSIM

Frame insertion and deletion are common types of inter-frame forgeries in digital videos. In this article, an efficient method based on consistency of quotients of mean of structural similarity ( MSSIM) was proposed. For a continuous shooting video, due to the continuity of video content, the quotient of MSSIM between consecutive frames is continuous. While for a video which has been tampered by inserting a certain number of frames from another video or deleting a certain number of frames, at the tamper point, the quotient of MSSIM becomes rather low, the continuity will be broken. According to the phenomena, MSSIM between every two adjacent frames iscalculated firstly. Then distinguishing feature is obtained by calculating the quotients between every two sequential MSSIMs. At last, insertion and dele-tion detection and localization can be achieved by using Tchebycheff inequality twice followed by deci-sion-thresholding. Experiments show that the proposed method has high detection accuracy against both frame insertion and deletion.