基于MROGH特征描述子的图像局部拷贝—旋转—移动行为取证

Copy-rotation-move forgery detection using the MROGH descriptor

各种数字图像编辑软件的普及让人们的生活更加多姿多彩,但同时也带来了更多的安全隐患——恶意的篡改者伪造图像、传播虚假内容的门槛极大地降低了。图像局部拷贝—移动(copy-move)是最为容易操作且有效的图像篡改手段之一,针对这种图像篡改手段的取证技术也在近年来受到了广泛关注。在图像的局部拷贝—移动篡改过程中,通常会包含一定程度的旋转操作,因此取证手段对于旋转的鲁棒性非常重要。本文提出了以筛选后的哈里斯角点结合MROGH(Multi-support Region Order-based Gradient Histogram)特征描述子的图像局部拷贝—旋转—移动篡改取证方法。实验证明,该方法具有良好的特征点覆盖率和旋转鲁棒性。

With different kinds of image editing software, it is getting much easier for an attacker to tamper a digital image. Thanks to its simplicity and effectiveness, copy-move forgery has become a commonly used approach to forging images and detection of this behavior has attracted much attention in recent years. Usually integrated in the copy-move forgery process is the rotation operation, which makes rotation invariance a critical property of the copy-move forgery detection methods. This paper proposes using screened Harris Corner Detector and the MROGH descriptor as feature detector and descriptor respectively, to gain better feature coverage and robustness against rotation. Experimental results illustrate the efficacy of the proposed method.