稳健性图像复制-黏贴篡改被动认证

图像复制-黏贴(copy-move)是一类的常见的图像篡改手段，篡改通过将图像中一部分区域复制并黏贴到同一幅图像另一区域后起到掩盖被覆盖内容的目的。由于篡改者为了使篡改更加逼真或者试图增加检测难度，往往在黏贴图像块之前对图像块进行加噪、模糊或者旋转缩放等后续处理。目前检测这类篡改的认证方法主要归纳为三类：变换域鲁棒特征子块匹配方法、旋转不变特征子块匹配方法和特征点匹配方法。本文对采用这三类方法的国内外文献进行了系统的分析和归纳并对未来研究方向进行了展望。

A study on approaches for detecting copy-move forgeries

Image copy-move modification is one kind of common tampering manipulations. The forgers duplicate some part of an image and move to another place in the same image. The simplest approach to detect this kind of forgery is dividing the image into blocks and exhaustively searching the identical pairs according to their pixel values. However, to make convincing forgeries, the duplicated regions are often retouched with illumination, geometrical and sharpness adjustments. We roughly classified copy-move detection methods into three categories: 1) robust block-features in frequency domain based methods; 2) block-features of rotation invariant based methods; 3) feature-points based methods. In this paper we outline most typical methods and summarize the technological progress in copy-move detection. Some new trends are also discussed in the conclusion.