Passive Image Copy–Move Forgery Detection Based on ORB Features

Digital image acquisition is extremely easy nowadays, but this convenience comes with the illegal purposes of criminals. Therefore, it is very necessary to identify the authenticity of the image. This paper proposes an adaptive image copy–move forgery detection method based on ORB features. Firstly, the proposed method performs an adaptive image super-pixel segmentation. Secondly, adaptive image sampling is carried out for each super-pixel block according to the block size. Third, quad-tree is used to store the key-points extracted by oFAST, and then the extracted key-points are described in rBRIEF. After that, Brute Force Matcher and hamming distance are adopted in the proposed method to perform key-points matching. Finally, MorphSnake algorithm is employed to locate the forged areas. Experiments have shown that compared with the current advanced detection methods, this one has stronger robustness to the attack of brightness, color, and contrast, and shorter running time.