Improved detection for copy-move forgery with multi-scale sliding windows

This work proposes an improvement solution for detecting copy-move forgery in individual image. The proposed method can significantly remove a large number of uncombined image block by using multi-scale sliding windows, so, it is different from the existing schemes. The proposed scheme consists of feature extraction, feature matching and uncombined block removing. DCT coefficient are extracted to design low-dimensional features which are considered to be more sensitive to copy region of image. Furthermore, we use fast k-means method to integrate image blocks with similar features as different clusters. Finally, multi-scale sliding windows is designed to further remove the uncombined blocks in detection results. Extensive experiments show that the proposed scheme is effective and efficient and outperforms the state-of-the-art copy-move detection methods with a lower false positive rate.