一种缩放情况下区分原始区域和篡改的方法

Method of Distinguishing Original Area from Duplicated Areas in Case of Scaling

为了在检测结果中区分出原始区域和粘贴区域,提出一种基于尺度不变特征转换(SIFT)和重采样痕迹的图像盲检测算法。该算法首先将图像分成互不重叠块,利用SIFT算法找出每个块中的关键点,对关键点匹配;然后在每个块中以匹配关键点的中心为种子进行区域生长得到篡改区域;最后分析每个图像块的频谱图,计算出缩放因子,区分原始区域和复制区域。实验结果表明,该算法简便快捷,在能够对篡改区域进行准确定位的同时,并能区分原始区域和复制区域。

In order to distinguish the original area from duplicated areas,a method based on scale invariant feature transform （SIFT）descriptors and rescampling is proposed. Firstly, the algorithm divides the image with non-overlapping blocks, uses the SIFT algorithm to find the key points and matches the key points. Then, original region and tampered area are obtained after region growing with the seeds of the center of the matched key points. Finally, the original region and tampered areas are exposed by means of comparing the Spectrogram of each block and the scal.ing factor. Experiments results show that the method is easily implemented. It can distinguish original region with tampered areas as well as locating the tampered areas accurately.