Duplicate Frame Detection in Forged Videos Using Sequence Matching

Abstract

This paper presents a method that can detect multiple frame duplication in forged videos. In this kind of forgery, a number of frames are copied and placed somewhere in a video for hiding or highlighting some incidents and/or objects. It becomes very challenging to identify the duplicated frames in such digital videos, as it is difficult to detect by the human eyes as well as by computer program. In past works, various methods have been proposed so far to detect the same. Despite that, there are very a few methods which can determine frame duplication in the video with both static and non-static background. To this end, this work proposes a simple yet effective method to detect the frame duplication forgery found in the videos. In this method, the structural similarity index measure (SSIM) is used to assess the resemblance between the consecutive frames. Finally, to detect and localize tampered frames, a searching algorithm is used. For experimental need, we have collected original videos from Urban Tracker, derf’s collection and REWIND databases, and then prepared the forged videos using frame duplication. Experimental results on these videos confirm that the present method detects such forgery with an impressive average frame duplication detection accuracy of 98.90%.