| 不同图像之间拼接篡改的数字图像检测研究 |
| |
| 引用本文： | 王丽侠.不同图像之间拼接篡改的数字图像检测研究[J].唐山学院学报,2018,31(6):18-26. |
| |
| 作者姓名： | 王丽侠 |
| |
| 作者单位： | 唐山学院 智能与信息工程学院, 河北 唐山 063000 |
| |
| 摘 要： | 对基于CFA（颜色滤波阵列）模型的篡改检测算法进行了改进。其检测过程为：首先，利用插值算法得到像素位置的预测误差，根据预测误差计算出CFA单元特征；然后，利用EM（期望最大化）算法估计特征模型的参数，算法对篡改位置的均值不做事先确定（从实验来看这种改进具有较好的效果）；最后，利用贝叶斯理论计算出每个像素点的似然率，根据似然率的不同来定位篡改区域。在进行单CFA阵列模式检测的情况下，对多种CFA阵列模式的图像也进行了检测分析，实验结果显示，该算法能够对多种CFA阵列模式的图像准确定位篡改区域。
|
| 关 键 词： | 图像检测 拼接篡改 颜色滤波阵列模型 期望最大化算法 |
Research on Detection of Digital Stitched-and-Tampered Images |
| |
| WANG Li-xia.Research on Detection of Digital Stitched-and-Tampered Images[J].Journal of Tangshan College,2018,31(6):18-26. |
| |
| Authors: | WANG Li-xia |
| |
| Institution: | School of Intelligence and Information Engineering, Tangshan University, Tangshan 063000, China |
| |
| Abstract: | The tampering detection method based on CFA Model is improved in this paper. The process is as follows:Firstly,to obtain the prediction error of the pixel position by using interpolation algorithm and calculate CFA unit features according to the prediction error; then, to estimate the parameters of the feature model with the EM algorithm and not to determine the average of tampering location beforehand(this improvement has a better effect in the experiment); finally,to calculate the likelihood ratio of each pixel with Bayesian Theory and to locate the tampered area according to the difference in the likelihood ratio. After the detection of a single CFA array pattern,the images with multifarious CFA array pattern are also detected. Experiment results show that tampered areas of images with multifarious CFA array pattern can be accurately located. |
| |
| Keywords: | image detection stitch and tamper CFA model EM algorithm |
|
| 点击此处可从《唐山学院学报》浏览原始摘要信息 |
| 点击此处可从《唐山学院学报》下载免费的PDF全文 |
