摘要
复制粘贴是图像篡改的常用手段,经典伪造检测方法将所有重叠图像块作为检测区域,算法时间复杂度高,邻近区域误检率大。为解决以上问题,提出将扩展Harris角点作为检测区域以降低算法复杂度。由于图像经过复制粘贴检测后往往会进行模糊、添加杂色、色彩调整等后处理,使得图像质量下降,本文结合NR图像质量评价给出更为准确的检测结果。实验证明本文算法对经过模糊、添加杂色、JPEG压缩等后处理的复制粘贴图像检测效率高,检测效果好。
Copy-move is a main method of image tampering. Many proposed copy-move forgery detection methods take all the overlapping image blocks as detection regions, so that the time complexity and the false detection rate of adjacent areas are high. In this paper, we use Harris corners as detection regions to decrease computational cost. Because .a forgery copy-move image undergoes a serial of image processing such as fuzzing, adding noise, JPEG compression, image rotation, etc. so the image quality always decrease. We combine harris corner extraction and NR quality assessment to achieve more accurate detection result. Experimental results show our method is robust and effective for the tampered images after post processing.
出处
《电子设计工程》
2014年第17期189-193,共5页
Electronic Design Engineering
