The Research On Methods For Copy-move Regions Forgery Detection In Digital Videos

Nowadays,the function of video processing tools becomes more and more powerful,editing and modifying video has been so easy even for common people.As a result,authenticating the content of digital image and video and detecting if it has been tampered becomes an important problem in judicial domain.However,existing video region-level copy-move forgery detection approaches suffer from shortcomings,such as low detection precision,poor robustness and high computation complexity.This paper make a deep study on two types of video region-level copy-move forgery,and propose two different forgery detection algorithms respectively.As existing video region-level inter-frame copy-move forgery detection methods often suffer from low detection precision,a spatial-temporal slice based detection algorithm is proposed in this paper.Firstly,divide video into groups and frame alignment is done in each group.Secondly,extract spatial-temporal slice from video,collect all pixels that having abnormal correlation and global correlation abnormal pattern is computed.Lastly,forged region is extracted from each tampered slice.The forgery region in each tampered frame is detected based on the detection result in all slices.Comparison experiment shows that,proposed algorithm not only has quite strong robustness,the accuracy and recall rate is 93.2%and 80.8%respectively as well.The detection result of our method improved significantly compared to previous algorithms.As existing video region-level intra-frame copy-move forgery detection methods often suffer from low detection precision and is time-consuming in algorithm execution,a SURF and optical flow based detection algorithm is proposed in this paper.Firstly,extract a set of keyframes from video based on keyframe selection algorithm.Secondly,detect and locate forged regions from keyframe based on SURF matching keypoints.Lastly,for the detection in non-keyframe,the sparse optical flow field is computed for matched keypoints and transform these keypoints to locate the suspicious tampered regions in non-keyframe.Further verify operation is done to ensure the correctness of detection result.Comparison experiment shows that,proposed algorithm not only has quite strong robustness,the accuracy and recall rate is 96.1%and 92.2%respectively as well.The detection result of our method improved significantly compared to previous algorithms and proposed algorithm increase 35%in detection efficiency.This paper study on digital video forgery detection technique,and mainly focus on video region-level copy-move forgery detection.For two different forgery types,two algorithms is proposed to detect and locate tampered regions in forged videos.Using a combination of theory and practice mode,for digital video forensics technology has carried on the beneficial exploration.