Research On Digital Video Copy-Move Forgery Detection

The rapid growth of portable and high resolution digital multimedia devices has increased the possibility of generating videos without any time and location. In addition, the advanced video-editing software has made people tampering videos with little effort. When a large number of edited or forged digital videos appear on websites or use to support legal evidences, it will have negative influence on the society. Therefore, verifying the authenticity and integrity of digital videos and detecting traces of tampering without requiring any per-embedded information have become an important research issue.There are many ways to tamper digital videos currently, copy-move forgery can better change the content and meaning of videos among of them. Copy-move forgery can be classified into two types:intra-frame tampering and inter-frame tampering. This way not only can make the frames introduce new objects or cover existed objects, but also can make the scenes occurred ahead or delay. Therefore, it will be more meaningful to detect the copy-move forgery.This paper studies the characteristics of copy-move forged videos, then the two new algorithms are proposed to detect the forged videos effectively. The main work of this paper can be summarized as follows:1. A video intra-frame copy-move forgery detection algorithm by fusing multiple features is proposed. The proposed algorithm by using block matching and3D filtering(BM3D) algorithm and2-D phase congruency to pre-processing video, and then extracts the first-order statistical characteristics of video frames, by using the differences between original frames and forged frames to detect tampering. Experimental results demonstrated that the proposed algorithm can detect video of intra-frame copy-move forgery effectively. Meanwhile, the new detection algorithm is better than the previous methods with respect to the dimensions of characteristic and precision rate.2. A video inter-frame copy-move forgery detection algorithm based on structural similarity is proposed. In this algorithm, the structural similarity is extended to measure the similarity between two frames of a video. Then a full-length video sequence is divided into some short overlapping sub-sequences. A temporal similarity measurement strategy between short sub-sequences is put forward to detect copy-move forgery. In addition, we can obtain an accurate forgery localization. Extensive experimental results show that the precision of this algorithm is higher than a previous relevant study.