A fast and accurate algorithm for copy-move forgery detection

In recent years, and with the presence of many efficient image processing tools, digital image forgery has become a serious social issue. Copy-move forgery is one of the most widely used methods for image forgeries in which a part of the image is copied and then pasted to another location in the same image. This procedure is usually used to add or cover a critical part of the image. In this paper, we propose a new fast and accurate algorithm for copy-move forgery detection in digital images. In the proposed algorithm, the image to analyze is segmented into overlapping square blocks with a predefined side length, each one of the blocks is split into equally spaced k sub-blocks. The sum of pixel intensities of each sub-block is used to form a k-dimensional vector with the help of sliding window and such vector is used as a feature for each block. The resulting features of all blocks are stored in a KD-tree. The block corresponding to each node in the KD-tree is checked with the block corresponding to the nearest neighbor of this node. If the correlation between such blocks is above a prespecified threshold, the two blocks are considered as clones. Experimental results and comparisons with a state of the art method show that the proposed algorithm is fast and accurate.