摘要
为了提高拼接篡改图像的检测准确率,利用视觉注意模型提出了一种新的图像拼接篡改盲鉴别算法。首先,采用改进的基于OSF的非线性滤波方法提取图像的边缘信息,得到边缘显著图像ECM;其次,利用视觉注意模型提取ECM的注意点,并采用显著边缘定位法锁定图像显著边缘处注意点,进而获取图像关键特征片段;接着,提取图像片段的Cr通道,并计算其小波重构图像;然后,针对小波重构图像,提取其扩展的DCT域的HMM特征,并采用SVM-RFE算法对所提取特征进行降维处理;最后,根据得到的特征向量,利用SVM对特征值进行训练并建立分类模型,从而实现自然图像和拼接篡改图像的分类识别。实验结果表明,针对哥伦比亚大学拼接篡改图像库,本文算法的正确检测率为96.32%。
In order to improve the detection accuracy of spliced images, a new blind detection based on the Visual Attention Model （VAM） was proposed in this study. First, the Edge Conspicuous Map （ECM） is created by an improved Order Statistics Filter （OSF） based nonlinear filtering approach; then, the ECM fixations are extracted by VAM, and the fixations on the boundaries are located by conspicuous edge positioning method, accordingly the key feature fragments are captured. Second, the Extended Hidden Markov Model （E-HMM） features are extracted from each wavelet reconstructed image of Cr channel of the fragments, and their dimensions are reduced by SVM-RFE. Finally, the above features are trained and classified using SVM, by which the spliced images can be identified from the natural ones. The experimental results show that, when testing on the Columbia image splicing detection dataset, the detection accuracy of the proposed method is 96.32%.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2014年第2期446-453,共8页
Journal of Jilin University:Engineering and Technology Edition
基金
国家青年科学基金项目(61305046)
吉林省青年科学基金项目(20130522117JH)
关键词
计算机应用
盲鉴别
图像拼接
视觉注意模型
扩展的隐马尔可夫模型
computer application
blind identification
image splicing
visual attention model
extended hidden Markov model
