摘要
现有图像来源辨识算法很少讨论测试图像在受到轻微图像处理后算法检测准确率的变化。以平板扫描仪所获得的数字图像为研究对象,提出一种基于特征的鲁棒源扫描仪辨识技术。根据不同扫描仪成像色纯、制作工艺的差异,提取反映以上差异的特征,包括颜色特征、图像质量特征以及邻域特征;然后利用支持向量机训练分类器用以辨识扫描仪的品牌/型号;最后对源扫描仪辨识算法的性能和鲁棒性进行了分析。实验结果表明:本文算法能优于前面提出的算法,并且具有较好的鲁棒性。
When testing images were subjected to minor image processing the change of detection rates was rarely discussed in the current image origin identification problem.A robust source scanner identification method was proposed,in this paper,by using acquired digital images of flatbed desktop scanners.First,the feature of color,image quality and neighborhood were proposed based on the difference of imaging color-purity and fabrication process of different scanner brands/models.Then,Support Vector Machine classifier was used to verifying the performance and robustness of our source scanner identification algorithm.Extensive experimental results demonstrate that the proposed method can effectively identify scanner brands/models and has better robustness than the previously.
出处
《山东大学学报（工学版）》
CAS
北大核心
2011年第2期62-65,74,共5页
Journal of Shandong University（Engineering Science）
基金
国家自然科学基金资助项目(60772115
60572140)
关键词
数字图像取证
图像来源
源扫描仪辨识
支持向量机
鲁棒性
digital image forensics
image origin
source scanner identification
support vector machine
robustness
