An Efficient Copy-Move Detection Algorithm Based on Superpixel Segmentation and Harris Key-Points

Abstract

Region duplication is a commonly used operation in digital image processing. Since region duplication could be utilized to easily tamper the raw content by intentional attackers, it has become a very important topic in image forensics. Most of the existing detection methods designed to region duplication are based on the exhaustive block-matching of image pixels or transformed coefficients. They may be not efficient when the duplicate regions are relatively smooth, or processed by some geometrical transformations. This has motivated us to propose a reliable copy-move forgery detection algorithm based on super-pixel segmentation and Harris key-points to improve the detection accuracy due to these specified attacks. For a given image, the proposed method first uses SLIC super-pixel segmentation and cluster analysis technique to partition the image content into complex regions and smooth regions. Then, a region description method based on sector mean is introduced to represent the relatively small image regions around each Harris point by adopting a well-designed feature vector. Thereafter, for both complex regions and smooth regions, we perform the feature matching operation, which is finally exploited to locate the tampered region. Experimental results have shown that, our algorithm significantly outperform some related works in terms of the detection accuracy when the test images are processed by blurring, adding noise, JPEG compression and rotating, which has shown the superior of our work.