Inter-frame forgery detection for static-background video based on MVP consistency

Abstract

Frame deletion and duplication are common inter-frame tampering methods in digital videos. In this paper, an efficient forensic method based on motion vector pyramid (MVP) and its variation factor (VF) is proposed to detect frame deletion and duplication in videos with static background. This method is composed of two parts: feature extraction and discontinuity point detection. In the stage of feature extraction, each frame of the video is transformed to grayscale image firstly. Then, motion vector pyramid (MVP) sequence and its corresponding variation factor (VF) are calculated for every two adjacent frames. In the stage of discontinuity point detection, forgery type is identified and tampering point is localized by performing modified generalized ESD test. Experimental results show that the proposed method is efficient at forgery identification and localization. Compared with other existing methods on inter-frame forgery detection, our proposed method is more generic.