Video motion forgery detection using motion residual and object tracking

Due to the extensive application of the video editing tools, video authentification has become an interesting topic. Video motion forgery is one of the most significant manipulations which alter the video sequence both temporally and spatially. We suggest an intra-frame motion forgery detection algorithm in this paper that detects the motion forgery throughout the video sequence. The motion residuals are applied to extract the moving parts. The moving objects are then analyzed to identify the similar ones. Then, the moving objects are tracked throughout the video sequence using the meanshift algorithm to determine the motion sequences. The two motion sequences are selected as matched if both the objects and their displacements are similar in the consecutive frames. By matching the motion sequences, the forged ones are determined. Various simulations have been conducted in different datasets to investigate the performance of the proposed scheme. We test video sequences at both frame and pixel levels and detect temporal and spatial motion forgery. For the frame level, the F 1 score of the proposed method on both the datasets of SULFA and GRIP was 89 % . Also, at the pixel level, the F 1 scores of the proposed method for the SULFA and GRIP datasets are 89 % and 81 % , respectively. The simulation results confirm the efficiency of the suggested method in locating the forged motion sequences and outperforming its rivals.