Identifying Video Tampering through Optical Flow Analysis: A Focus on Frame Manipulations

This research addresses the critical problem of video tampering detection, focusing on frame deletions, insertions, and duplications. An unconventional approach is proposed, utilizing Farneback optical flow to detect inconsistencies in motion patterns indicative of tampering. To enhance the accuracy of optical flow computation, Contrast-Limited Adaptive Histogram Equalization (CLAHE) is applied to manage artifacts resulting from compression, lighting changes, and occlusions. Additionally, temporal filtering techniques are employed to reduce noise sensitivity. Our methodology includes evaluating motion consistency across frames, extracting texture features from the flow field, and applying thresholding techniques to identify abnormal motion characteristics. Temporal consistency checks are performed to ensure the persistence of detected forged regions over multiple frames. Experimental results on publicly available datasets demonstrate the effectiveness of our approach, achieving a TPR of 93.15% and a FAR of 6.65%, highlighting its high accuracy and robustness in identifying video tampering.