-
PS技术的发展使得伪图像现象层出不穷,这对人们信息的获取造成了一定的负面影响[1-2].对此,人们提出很多方法来检测图像中篡改的内容.基于图像块检测的方法是一种常用的图像伪造检测方法,如周萌萌等人[3]对计算所得的亮度图像以及局部亮度变化图像进行图像分块,并计算每个图像块对应的共轭不变矩,利用共轭不变矩特征值对图像块进行分组,接着对临近组的图像块进行相似值测量,进而检测出伪造内容,实验表明,该方法能对伪造图像进行检测,而且检测速度很快,但是对于经过旋转后的伪造图像效果不佳.对此,Ghulam等人[4]利用可控金字塔获取特征点,再利用局部二值模式完成检测,实验表明,该方法对弥补前面的检测缺陷具有一定的效果,但是检测结果存在一定的错误.又如Zheng等人[5]利用SIFT算法良好的仿射不变性能,提出了一种基于块联合特征点的图像伪造检测算法,进而完成图像伪造检测,实验表明,该方法具有很好的鲁棒性能,对伪造图像检测效果良好,但是检测结果存在漏检测现象.朱叶等人[6]利用图像DOG区域提取方法检测图像的特征点,接着利用MIOP方法生成特征描述子,通过欧氏距离比值的方法进行特征匹配,进而用RANSAC方法去除误匹配,精确定位伪造图像,实验表明,该方法检测效果较好,而且具有较好的鲁棒性能,但是该方法生成的特征描述子维度太高,导致算法的计算量太大.
对此,本文提出了一种基于双重特征匹配耦合Hough变换聚类的图像伪造检测算法.首先,引入FAST算法对图像进行特征检测,利用特征点的梯度特征,通过梯度直方图统计法以及求取同心圆梯度特征分别获取特征点的主方向以及特征向量,生成较低维度的特征描述子;然后,将特征点的HSI颜色分量以及特征点的特征向量作为特征点的双重特征,利用双重特征匹配法则,进行特征点的匹配;接着,通过Hough变换对匹配后的特征点进行聚类,完成伪造内容的检测和定位.最后,测试该算法的检测性能.
Image Forgery Detection Algorithm Based on FAST Operator and Multi-feature Matching
-
摘要: 当前图像伪造检测算法大多采用最近邻与次近邻比值法进行特征匹配来完成图像伪造检测,存在较多的错误检测以及漏检测现象,基于此提出了一种基于FAST算子与多特征匹配的图像伪造检测算法.首先,基于FAST算法与Bresenham方法,构造以像素点为中心的圆形区域,提取图像特征;然后,通过梯度直方图统计法判定特征点的主方向,以特征点为中心建立两级同心圆,并通过求取同心圆在指定方向上的梯度特征,生成特征向量和特征描述子;最后,提取特征点的HSI颜色分量,将HSI颜色分量以及特征点的特征向量作为双重特征,设计了双重特征匹配法则,实现特征匹配.引入Hough变换,对匹配特征点进行聚类,定位伪造内容.实验结果显示,与当前图像匹配算法相比,所提算法具有更高的检测正确度与鲁棒性能.Abstract: In order to solve the current image forgery detection algorithms, the nearest neighbor and nearest neighbor ratio method is used to perform image forgery detection, which results in more error detection and leakage detection. An image forgery detection algorithm based on dual feature matching coupled Hough transform clustering has been proposed in this paper. Firstly, the FAST method is used to construct the circular region centered on the pixels, and the image features are extracted by Bresenham method. Secondly, through the gradient histogram statistics method to determine the main direction of the characteristic points, the feature points as the center to construct two concentric circles, and through calculating the concentric gradient in the specified direction, generates a feature vector generation feature descriptor. And Lastly, the HSI color components of feature points are extracted, and the HSI color components and feature vectors are used as the dual features of feature points and then dual feature matching rules are formulated to realize feature matching. The Hough transform is introduced to cluster the matching feature points to locate the fake content. The results show that this algorithm has higher detection accuracy and better robust performance compared with the current image matching algorithm.
-
[1] 张晓琪, 侯世英.基于圆形均分法耦合双重制约的图像伪造检测算法[J].西南师范大学学报(自然科学版), 2018, 43(1):47-53. doi: http://www.xnsfdxxb.cn/index.php?m=content&c=index&a=show&catid=122&id=3088 [2] 柴建伟, 刘婷.改进的SIFT耦合特征点集群的图像伪造检测算法[J].西南师范大学学报(自然科学版), 2018, 43(3):34-41. doi: http://www.cnki.com.cn/Article/CJFDTotal-XNZK201803006.htm [3] 周萌萌, 杨鸿波, 高晶敏, 等.基于快速分块的复制-粘贴图像检测算法[J].计算机应用, 2016, 36(S1):117-121. doi: http://d.old.wanfangdata.com.cn/Periodical/jsjyy2016z1030 [4] MUHAMMAD G, MUNNER H A, MUHAMMAD H.Image Forgery Detection Using Steerable Pyramid Transform and Local Binary Pattern[J]. Machine Vision and Applications, 2014, 25(4):985-995. doi: 10.1007/s00138-013-0547-4 [5] ZHENG J B, LIU Y N, REN J C, et al.Fusion of Block and Keypoints Based Approaches for Effective Copy-Move Image Forgery Detection[J]. Multidimensional Systems and Signal Processing, 2016, 27(4):989-1005. doi: 10.1007/s11045-016-0416-1 [6] 朱叶, 申铉京, 陈海鹏.基于混合灰度序模式的图像复制-粘贴篡改盲鉴别算法[J].吉林大学学报(工学版), 2017, 47(4):1280-1285. doi: http://d.old.wanfangdata.com.cn/Periodical/jlgydxzrkxxb201704038 [7] 周莉莉, 姜枫.基于FAST和BRIEF的图像匹配算法[J].计算机工程与设计, 2015, 36(5):1269-1273. doi: http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=jsjgcysj201505030 [8] IVANOV V A, KIRICHUK V S, KOSYKH V P.Specific Features of Detecting Point Objects in Images Formed by a Detector Array[J]. Optoelectronics, Instrumentation and Data Processing, 2016, 52(2):113-120. doi: 10.3103/S8756699016020023 [9] 安维胜, 余让明, 伍玉铃.基于FAST和SURF的图像配准算法[J].计算机工程, 2015, 41(10):232-235, 239. doi: http://d.old.wanfangdata.com.cn/Periodical/jsjgc201510043 [10] BENRHOUMA O, HERMASSI H, ABD EL-LATIF A A, et al. Chaotic Watermark for Blind Forgery Detection in Images[J]. Multimedia Tools and Applications, 2016, 75(14):8695-8718. doi: 10.1007/s11042-015-2786-z [11] 张瑞芳, 程晓辉, 宋子航, 等.融合灰色马尔科夫理论的二进小波图像的复制-粘贴篡改检测算法[J].桂林理工大学学报, 2014, 34(4):775-781. doi: 10.3969/j.issn.1674-9057.2014.04.030 [12] doi: http://d.old.wanfangdata.com.cn/OAPaper/oai_doaj-articles_81e780e5e6f7473a1b4d676b7b6fa549 SWAPNIL H K, AVINASH D G.Copy- Move Attack Forgery Detection by Using SIFT[J]. International Journal of Innovative Technology and Exploring Engineering, 2013, 5(2):221-224. [13] 李晓飞, 李鹏飞.基于SIFT的伪造图像盲检测算法[J].长春大学学报, 2014, 24(10):1354-1357. doi: http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=ccdxxb201405011 [14] doi: http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=aad98773454c0f2c1a292d2624132331 LIANG Y R, LIU G P, ZHOU N R, et al.Color Image Encryption Combining a Reality-Preserving Fractional DCT with Chaotic Mapping in HSI Space[J]. Multimedia Tools & Applications, 2016, 75(11):6605-6620. [15] ALAHMADI A, HUSSAIN M, ABOALSAMH H.Passive Detection of Image Forgery Using DCT and Local Binary Pattern[J]. Signal, Image and Video Processing, 2017, 11(1):81-88. doi: 10.1007/s11760-016-0899-0 [16] 李宁, 许树成, 邓中亮.基于HSI色彩坐标相似度的彩色图像分割方法[J].现代电子技术, 2017, 40(2):30-33, 38. doi: http://d.old.wanfangdata.com.cn/Periodical/xddzjs201702007 [17] 徐庆嵩, 尹立新, 张伟伟.一种使用HSI颜色空间改进的图像修复算法[J].计算机应用与软件, 2015, 32(6):216-219, 248. doi: 10.3969/j.issn.1000-386x.2015.06.053 [18] 龚一珉, 孙刘杰.基于Hough变换旋转校正的QR码全息水印算法[J].包装工程, 2017, 38(9):213-218. doi: http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=bzgc201709040 [19] MASSSONE A M, BELTRAMETTI M C, RICCA G.Detecting Curves of Symmetry in Images Via Hough Transform[J]. Mathematics in Computer Science, 2016, 10(1):179-205. doi: 10.1007/s11786-016-0245-5 [20] CALATRONI L, VAN-GENNIP Y, BIBIANE-SCHÖNLIEB C.Graph Clustering, Variational Image Segmentation Methods and Hough Transform Scale Detection for Object Measurement in Images[J]. Journal of Mathematical Imaging & Vision, 2017, 57(2):269-291. [21] 曾硕, 王玲, 李昆明.一种基于Zernike不变矩和改进的Hough变换的图像检索方法[J].计算机应用与软件, 2015, 32(12):210-213. doi: 10.3969/j.issn.1000-386x.2015.12.049 [22] doi: http://d.old.wanfangdata.com.cn/NSTLHY/NSTL_HYCC0214453220/ FARUKH M, HAMBARDE A R, KESKAR A G.Copy Move Forgery Detection using DWT and SIFT features[J]. Intelligent Systems Design and Applications, 2014, 34(10):188-193. [23] 李景富, 张飞.凸优化耦合传感器模式噪声的图像伪造检测[J].计算机测量与控制, 2015, 23(5):1678-1681, 1685. doi: http://d.old.wanfangdata.com.cn/Periodical/jsjzdclykz201505069 [24] 李伟征, 白光伟, 涂秀梅.于局部相位量化的图像伪造检测算法[J].计算机工程与设计, 2017, 38(2):449-452, 459. doi: http://d.old.wanfangdata.com.cn/Periodical/jsjgcysj201702032 [25] BIRAJDARG K, MANKAR V H.Blind Method for Rescaling Detection and Rescale Factor Estimation in Digital Images Using Periodic Properties of Interpolation[J]. AEU-International Journal of Electronics and Communications, 2014, 68(7):644-652. doi: 10.1016/j.aeue.2014.01.013