| 一种新颖的数字图像篡改检测方法 |
| |
| 引用本文： | 王泓鑫,李 硕.一种新颖的数字图像篡改检测方法[J].教育技术导刊,2009,19(8):221-225. |
| |
| 作者姓名： | 王泓鑫 李 硕 |
| |
| 作者单位： | 大连科技学院 软件技术学院，辽宁 大连 116052 |
| |
| 基金项目： | 国家自然科学基金项目（61701212）|中国博士后科学基金项目（2017M621135） |
| |
| 摘 要： | 检测数字图像复制—粘贴型篡改是目前研究热点之一。多数复制—粘贴篡改检测方案只对刚性平移篡改具有较好的检测精度，无法有效应对更复杂的几何变换和常规信号攻击。为优化数字图像复制—粘贴篡改检测效率和精度，首次提出一种使用极坐标复指数变换（PCET）与一致性敏感哈希（CSH）的高效检测算法。首先，计算滑动窗口 PCET 系数，将其作为不变的图像局部特征|然后，根据这些特征，利用 CSH 算法快速、精确地匹配大量密集分块|最后，使用基于密集线性滤波的后处理算法消除匹配结果中的错误匹配并定位重复区域，得到最终检测结果。根据实验结果可知，该算法不仅检测精度平均提高 12.06%，而且处理时间平均缩短315.64s。因此利用对几何变换和常规信号攻击鲁棒的 PCET 系数刻画图像局部特征，并采用基于图像一致性的快速高精度匹配算法 CSH，可有效优化复制—粘贴篡改检测精度和效率。
|
| 关 键 词： | 复制—粘贴型篡改 图像矩 区域复制 旋转不变 |
| 收稿时间： | 2019-12-13 |
Block-based Image Copy-move Forgery Detection Using PCET Feature and CSH Matching |
| |
| WANG Hong-xin,LI Shuo.Block-based Image Copy-move Forgery Detection Using PCET Feature and CSH Matching[J].Introduction of Educational Technology,2009,19(8):221-225. |
| |
| Authors: | WANG Hong-xin LI Shuo |
| |
| Institution: | School of Soft Technique，Dalian Institute of Science and Technology，Dalian 116052，China |
| |
| Abstract: | Detecting copy-move forgery of digital images is one of the current research hotspots. However，most existing copy-move forgery detection schemes are generally only robust to translation，and they are not effective in dealing with more complex geometric transformations and conventional signal processing attacks. In addition，such algorithms are often time-consuming and cannot meet the needs of real-time detection. In order to optimize the efficiency and accuracy of digital image copy-move forgery detection，a fast and effective algorithm using polar complex exponential transform（PCET）and coherency sensitive hashing（CSH）was proposed in this paper. First，the algorithm calculates the PCET coefficients of the sliding windows as invariant local features. Then，based on these feature vectors，a large number of dense windows will be quickly and accurately matched by the CSH algorithm. Finally，the dense linear fitting（DLF）based post-processing algorithm is used to eliminate the mismatches and locate the duplicated regions. In this paper，five challenging tampered images are selected from GRIP and FAU databases，and the accuracy and speed are compared with classic algorithm. According to the experimental results，the proposed algorithm not only improves the average detection accuracy by 12.06%，but also reduces the average processing time by 315.64 seconds. PCET coeffi-cients that are robust to geometric transformations and conventional signal processing attacks are used to characterize image local features. Moreover，image coherency based CSH is introduced for fast and high-precision feature matching. Therefore，the proposed algorithm is able to effectively optimize the accuracy and efficien? cy of detection. |
| |
| Keywords: | copy-move forgery image moment region duplication rotation invariance |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
