基于Harris兴趣点和空域均值信息的图像复制粘贴窜改检测算法

提出了一种基于兴趣点检测和特征匹配的图像复制粘贴窜改检测方法。首先采用Harris算子检测图像中的角点作为兴趣点, 然后提取以兴趣点为中心的邻域内空域的五个均值特征形成特征向量, 最后记录相等位移矢量的发生频率并通过阈值化处理得到匹配的兴趣点, 从而标志复制粘贴区域。仿真实验表明, 该算法不仅可以有效检测多区域复制粘贴窜改操作, 而且能够有效抵抗多种窜改后处理操作, 包括加性高斯白噪声, JPEG压缩, 对比度、亮度和曝光度调整以及JPEG压缩和加噪的混合操作。

Detection algorithm of image copy-move forgery based on Harris and spacial average information

This paper proposed a detection method of image copy-move forgery based on the interest points detection and feature matching. First, it applied Harris operator to detect corners as interest points, then extracted five spacial average features in the neighborhood of which as the center to constitute feature vector, and finally recorded the occurrence frequency of position transfer vector to obtain matching interest points with thresholding processing, thus marked copy-move regions. Experimental results show that the algorithm can not only detect multi-region copy-move manipulation effectively, but also be able to effectively resist various tampering post-processing operations, including additive white Gaussian noise, JPEG compression, contrast, brightness and exposure adjustment and the hybrid operation of JPEG compression and noise addition.