Detection Of Copy-move Forgery In JPEG Images Based On SURF And BAVM Factor

As the digital image processing software is increasingly powerful, digital image tampering is becoming more convenient and easier, and the authenticity of the digital image is gradually decreasing. So the digital image forensics, especially passive image forensics, has become a particularly important and hot research field nowadays. Copy-move forgery is the most common used image forgery, and JPEG format is the most popular image saved format, so research on forensics for detecting copy-move forgery in JPEG images is of great importance.This paper, first introduce the model and theory of copy-move forgery in JPEG images. Aiming at the problem that in existing methods that can’t distinguish the copied region from the tampered region, a novel method based on Blocking Artifact Characteristics Matrix (BACM) factor in JPEG images is proposed in this paper. The proposed method distinguishes the copied region from the tampered region by matched Block Artifact Grids (BGA) which is decided by the BACM factor.As the existing block-matching based methods are computationally complex and lack of robustness, to alleviate these problems, a method for automatic detection and location of the tampered region is proposed based on Speeded Up Robust Features (SURF) matching and BACM factor. Firstly, obtain image SURF keypoints and collect descriptor vectors at the detected keypoints. Secondly, match the detected keypoints by computing the Euclidean distance of their descriptor vectors. Next, use the matched SURF key-points to estimate the affine transform parameters. After discounting the estimated transforms, all pixels within the copy-move regions can be obtained. Finally, distinguish the copied region and the tampered region from the copy-move regions according to the BACM factors.The tested images are chosen from the UCID (Uncompressed Color Images Database) and McGill Calibrated color image database. The proposed method is implemented using Matlab. The experimental results indicate that our proposed method is effective for the detection and location of copy-move forgery in JPEG images.