基于相位相关的图像区域复制篡改检测与定位

Detection and location of image region duplication forgery based on phase correlation

针对区域复制篡改检测算法检测效率低问题,提出一种基于相位相关的检测与定位算法。算法首先对图像进行滑窗式分块,滑动步长为多个像素,通过对图像块灰度均值排序筛选出可能匹配的图像块;然后,利用相位相关的互功率谱脉冲函数峰值确定两个图像块是否匹配以及二者的相对位移;最后,通过统计匹配块对偏移距离方法消除误匹配,并对篡改区域进行标记。实验结果表明,该算法不仅检测效率高,而且对有损压缩、平滑滤波、噪声等常见的后处理操作具有鲁棒性。

Most existing algorithms for detecting region duplication forgery are of low efficiency. Concerning this problem, this paper proposed a detecting and locating algorithm based on phase correlation. Firstly, the image was divided into small blocks by the means of sliding window. It skipped muhi-pixels every time. Those blocks that had more matching possibility were chosen through sorting grayscale mean of block. Secondly, whether two blocks were matched was decided and their relative translation was calculated using the impulse function peak of the cross-power spectrum. Lastly, the mismatched blocks got removed according to the number of matched blocks that had same offset distance, and the forgery regions were marked according to the matched blocks. The experimental results show that the proposed algorithm is efficient and robust to such operations as lossy compression, blur filter and noise.