基于视频光流特征一致性的帧间篡改取证算法研究

发布时间：2018-01-05 01:20

本文关键词：基于视频光流特征一致性的帧间篡改取证算法研究　出处：《北京交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着视频信息的数字化、视频处理技术的不断进步以及国际互联网的普及,数字视频得以在世界范围内广泛传播与应用。在保证视频真实性的前提下,数字视频可以作为一种场景记录真实还原发生过的活动,因而广泛应用于法律、政治和科研等领域。然而,开放的互联网环境和便捷的信息存取以及处理技术也带来了许多数字视频的安全问题。数字视频取证技术作为一种有效的视频篡改检测技术,近十几年来得到了研究学者的广泛关注,成为信息安全领域的一个重要组成部分,并得到了广泛的研究和应用。本论文对视频的帧间篡改取证技术进行了探讨和分析,发现现有算法都是在已知视频是帧间篡改视频的情况下,对视频中的篡改点进行定位,缺少检测帧间篡改视频的相应算法,因此提出新的视频特征用于对原始视频和帧间篡改视频的分类检测,主要完成了以下三部分工作：1.论文深入分析了光流一致性特征应用于数字视频篡改检测的可行性理论依据。在没有被篡改的原始视频中,视频帧间的光流具有连续性,而当视频被帧间篡改后,在篡改点左右的两帧图像间的光流连续性被破坏。根据这一现象,论文将提取光流一致性作为数字视频帧间篡改检测的特征。2.论文提出了一种新的基于光流一致性特征的帧间篡改检测算法。该算法在一个较大的视频库中提取原始视频和篡改视频的光流一致性特征,经过归一化、量化和降维处理后,将一部分样本放入支持向量机进行训练,并使用另一部分样本测试分类结果。实验结果表明,该算法能够有效区分原始视频和帧间篡改视频。3.论文进一步将所提出的算法应用于不同种类的帧间篡改视频的区分,结果表明算法在区分帧插入篡改视频和帧删除篡改视频时仍然有效。论文算法在计算复杂度方面还有进一步提升的空间,有待在后续工作中进行改进。
[Abstract]:With the digitization of video information, the continuous progress of video processing technology and the popularity of the Internet, digital video can be widely spread and applied in the world, under the premise of ensuring the authenticity of video. Digital video can be used as a kind of scene to record the activities that actually happened, so it is widely used in the fields of law, politics, scientific research and so on. The open Internet environment and convenient information access and processing technology also bring many security problems of digital video. Digital video forensics technology as an effective video tamper detection technology. In the past ten years, it has been widely concerned by researchers, and has become an important part of the field of information security. And has been widely studied and applied. This paper discusses and analyzes the inter-frame tampering and forensics technology of video. It is found that the existing algorithms are all known that the video is inter-frame tampered with the video. To locate the tampering points in the video, there is no corresponding algorithm to detect the tampered video between frames, so a new video feature is proposed for the classification detection of the original video and the inter-frame tampered video. The thesis deeply analyzes the feasibility of the application of optical flow consistency features in digital video tamper detection. In the original video without tampering. The optical flow between video frames is continuous, and when the video is tampered with between frames, the continuity of the optical flow between the two frames around the tampering point is broken, according to this phenomenon. In this paper, optical flow consistency is used as the feature of digital video inter-frame tamper detection. A new algorithm of inter-frame tampering detection based on optical flow consistency is proposed. The algorithm is proposed in a large video library. The optical flow consistency features of the original video and the tampered video are obtained. After normalization, quantization and dimensionality reduction, some samples are trained by support vector machine, and the classification results are tested by another part of samples. The experimental results show that. This algorithm can effectively distinguish the original video from the inter-frame tampered video. The paper further applies the proposed algorithm to different kinds of inter-frame tampering video. The results show that the algorithm is still effective in distinguishing frame insertion tamper video from frame delete tamper video.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP309

【相似文献】