针对视频运动补偿帧率提升篡改的主动混噪取证算法  

Active Noised-mixed Forensics Algorithm for Tampering of Video Motion-compensated Frame Rate Up-conversion

作　　者：李然[1,2] 梅腊腊 邬长安 朱秀昌[2] 

机构地区：[1]信阳师范学院计算机与信息技术学院,信阳464000 [2]南京邮电大学通信与信息工程学院,南京210003

出　　处：《电子与信息学报》2018年第3期713-720,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61501393)~~

摘　　要：运动补偿帧率提升(MC-FRUC)是常见的视频时域篡改手段。现有方法依靠被动分析视频统计特征发现MC-FRUC篡改,然而,视频统计特性的非平稳性影响了取证性能的稳定性。该文提出一种主动混噪取证算法,通过预先混入统计特性已知的高斯白噪声,提高MC-FRUC取证的准确度。首先,利用伪随机序列生成高斯白噪声,加入原始视频序列。接着,由小波系数的绝对中位差预测各视频帧中混入高斯噪声的标准差。最后,检测高斯噪声标准差的时域变化周期性,通过硬阈值判决,自动甄别MC-FRUC篡改。实验结果表明,针对不同的MC-FRUC伪造方法,提出算法均表现出良好的取证性能,尤其是当采用去噪、压缩等操作后处理视频后,提出算法仍能确保较高的检测准确度。Motion-Compensated Frame Rate Up-Conversion（MC-FRUC） is one of the common temporal-domain tampering methods of video. The existing methods recognize MC-FRUC tampering by passively analyzing statistical characteristics of video; however, the non-stationarity in statistics of video affects the stability of forensics. This paper proposes an active noise-mixed forensics algorithm. First, white Gaussian noises are produced using a pseudorandom sequence, and these noises are added into the original video sequence. Second, based on the median absolute deviation of wavelet coefficients, the standard deviation of mixed Gaussian noises in each video frame is estimated. Last, the periodicity of standard deviation varying in time domain is detected, and MC-FRUC tampering with a hard-thresholding operation is automatically identified. Experimental results indicate that the proposed algorithm presents better performance of forensics for various MC-FRUC methods, and can still ensure high detection accuracy especially after videos are denoised or compressed.

关 键 词：运动补偿帧率提升 主动取证 高斯噪声 绝对中位差 周期性检测 

分 类 号：TN919.8[电子电信—通信与信息系统]