Article Preview
TopIntroduction
Using the powerful video editing software, video manipulation can be performed easily and the detection of tampered videos is difficult through human vision. To address the harmful impacts caused by video forgeries, video forensics has attracted wide attentions (Rocha, Scheirer, Boult, & Goldenstein, 2011; Milani, Fontani, Bestagini, Barni, Piva, Tagliasacchi, & Tubaro, 2012). Some inherent traces left by video editing operations can help the detection of video falsification such as the differential energy of residue (Hsu, Hung, Lin, & Hsu, 2008), the motion residual (Feng, Xu, Jia, Zhang, & Xu, 2016), and double compression artifacts (Jiang, Wang,Sun, Shi, & Wang, 2013).
Motion-Compensation Frame-Interpolation (MCFI) is another special frame based video manipulation, which periodically synthesizes intermediate frames to alleviate the motion discontinuity of low frame-rate videos (Yoo, Kang, & Kim, 2013; Li, Gan, Cui, Tang, & Zhu, 2014). Though MCFI is originally proposed to improve the visual quality or increase the frame-rate of low frame rate videos, it still might be used by a falsifier for malicious purposes. First, when faked frame-rate videos are released over video-sharing websites, they will not only waste many storage space but also mislead user’s visit. Second, two videos with different frame rates might be spliced by up-converting the low frame rate video to match the higher one. Third, MCFI might invalidate near-duplicate video detection or the video watermarking system because of the loss of temporal synchronism.
Some studies have been proposed to detect the use of MCFI. Estimating original frame-rate of faked videos was firstly proposed by using video-level artifacts such as prediction error (Bestagini, Battaglia, Milani, Tagliasacchi, & Tubaro, 2013), motion artifact (Jung, & Lee, 2018) and noise variation (Li, Liu, Zhang, Li, & Fu, 2018). They can obtain desirable results, yet cannot locate interpolation frames, let alone identify the adopted MCFI. Then, Yao et al. (Yao, Yang, Sun, & Li, 2016)and Xia et al. (Xia, Yang, Li, Li,& Sun, 2017) located the interpolated frames by the periodicity of edge-intensity and average texture variation, respectively. Subsequently, the localization problem of interpolated frames is discussed under real-world scenarios by employing Tchebichef moments (Ding, Zhu, Li, Li, & Yang, 2018). Recently, Yao et al. (Yao, Ni, & Zhao, 2019), as a pioneer, makes use of the MCFI strategy to invalidate inter-frame continuity based video forensics detection, and then present a global and local joint feature to attack this anti-forensic strategy. Besides, a detector is further proposed to judge the absence or presence of MCFI forgery in an environment of unknown MCFI techniques (Ding, Li, Xia, He, & Yang, 2019). As far as we know, there is only one work (Ding, Yang, Li, Zhang, Li, & Sun, 2017) about the identification issue of the adopted MCFIs, in which residual signal was firstly exploited as the tampering clue for identification of the adopted MCFIs. But, its performance deteriorates for videos encoded with relatively bigger quantization parameters.