Abstract: An efficient digital image tamper detection method is presented in this paper, which combines Discrete Wavelet Transform（DWT） with ORB feature extraction algorithm. It can meet the requirements of real-time feature extraction and copy-move tamper detection. Frist DWT is applied on an image to decompose it into LL, LH, HL and HH four parts. Then ORB algorithm is applied on LL part to extract the key points, binary descriptor vector—rBRIEF is constructed. Finally, to achieve the purpose of tamper detection, hamming distance is applied to match similar key points. The experimental results show that the method can effectively detect this type of tampering for images, and it shows robustness when tempered image receives translation, rotation, noise and synthesized attack. Simultaneously, the matching time and storage type are less than SIFT algorithm.

Key words: copy-move tempering, Discrete Wavelet Transform（DWT）, Oriented FAST and Rotated BRIEF（ORB）, feature extraction, tamper detection

摘要： 针对数字图像常见的区域复制篡改方式和实时特征提取要求，结合离散小波变换（DWT）和ORB特征提取算法，提出了一种高效的数字图像区域复制篡改检测方法。该方法首先使用DWT将图像分解为LL、LH、HL、HH四部分；然后将ORB算法作用于LL部分，提取关键点，构建二进制描述向量rBRIEF；最后使用汉明距离匹配相似关键点，达到检测图像篡改的目的。通过实验结果证明，该方法不仅可以有效地检测出区域复制篡改图像，同时对平移、旋转、噪音和综合攻击后的篡改图像具备一定的鲁棒性，且与SIFT等算法相比具有更短的检测时间和更少的存储空间。

关键词: copy-move篡改, 离散小波变换（DWT）, ORB方法, 特征提取, 篡改检测