使用 PSNR 周期特性检测视频帧率上转伪造

摘要

帧率上转（FRUC）是最常用的一种视频编辑技术，它在原始视频帧间周期性地插入新的帧，以便增加视频的帧率，这种技术经常用于两段不同帧率的视频拼接伪造中。为了减少视觉痕迹，高级的 FRUC 方法通常采用运动补偿的插值方式，这也带来了针对这种插值伪造检测的挑战。提出一种新的简单但有效的方法，可正确检测出这种伪造，并能估计出视频的原始帧率。该方法利用了 FRUC 算法生成的插值帧与相邻原始帧构成的视频序列再次插值重建得到的帧对在 PSNR 上的周期性差异。测试序列的实验结果表明该方法检测准确率高，其中对有损压缩视频序列的测试结果进一步证实了该方法的实际使用价值。%Frame-rate up-conversion (FRUC)refers to interpolating new frames between original video frames to increase the frame rate,which is widely used in video splicing forgery of two video with different frame rate.Advanced FRUC methods of-ten use motion compensated interpolators to reduce the visual artifacts,which brings challenges for this kind of forgery detec-tion.This paper proposed a simple yet effective detection method of this kind of forgery that estimated the original frame rate. This method used the periodic changes of PSNR for each pair frames,i.e.,the constructed interpolation frame by FRUC and the original frame.Experimental results on several sequences show that the proposed approach achieves desirable performance in terms of detection accuracy.It is also validated for those video sequences after PRUC and lossy compression,which further proves its robustness in a real application scenario.