Gabor特征和字典学习算法在打印文件鉴别中的应用
发布时间：2018-12-11 16:40
【摘要】：为了改善计算机打印文件的自动鉴别性能,提出了一种基于Gabor特征和Fisher判别准则核字典学习的激光打印文件鉴别算法。首先提取字符图像的Gabor幅值特征,同时将Gabor数据特征映射到高维核空间进行主成分分析;再将降维的特征作为初始字典,进行基于Fisher判别准则的字典学习;最后利用稀疏表示分类方法进行鉴别。在自建数据库上的实验结果表明Gabor特征在打印文件机源认证中是一种有效的鉴别特征,实验结果还验证了Gabor特征和Fisher判别准则核字典学习算法的有效性,打印文件源打印机正确鉴别率可达95.79%。
[Abstract]:In order to improve the automatic discrimination performance of computer printed files, a laser print file identification algorithm based on Gabor features and Fisher discriminant criterion kernel dictionary learning is proposed. Firstly, the Gabor amplitude feature of character image is extracted, at the same time, the feature of Gabor data is mapped to high dimensional kernel space for principal component analysis, then the reduced dimension feature is taken as the initial dictionary, and the dictionary learning based on Fisher discriminant criterion is carried out. Finally, the sparse representation classification method is used to identify. The experimental results on the self-built database show that the Gabor feature is an effective discriminant feature in the print file source authentication. The experimental results also verify the validity of the Gabor feature and the Fisher discriminant criterion kernel dictionary learning algorithm. The correct identification rate of print file source printer can reach 95.79.
【作者单位】： 湖北工程学院物理与电子信息工程学院;武汉大学电子信息学院;
【基金】：湖北省教育厅项目(B2015033) 湖北工程学院科研项目(201511) 湖北省大学生创新训练项目(201610528004)资助
【分类号】：TP301.6;TP334.8
本文编号：2372882
[Abstract]:In order to improve the automatic discrimination performance of computer printed files, a laser print file identification algorithm based on Gabor features and Fisher discriminant criterion kernel dictionary learning is proposed. Firstly, the Gabor amplitude feature of character image is extracted, at the same time, the feature of Gabor data is mapped to high dimensional kernel space for principal component analysis, then the reduced dimension feature is taken as the initial dictionary, and the dictionary learning based on Fisher discriminant criterion is carried out. Finally, the sparse representation classification method is used to identify. The experimental results on the self-built database show that the Gabor feature is an effective discriminant feature in the print file source authentication. The experimental results also verify the validity of the Gabor feature and the Fisher discriminant criterion kernel dictionary learning algorithm. The correct identification rate of print file source printer can reach 95.79.
【作者单位】： 湖北工程学院物理与电子信息工程学院;武汉大学电子信息学院;
【基金】：湖北省教育厅项目(B2015033) 湖北工程学院科研项目(201511) 湖北省大学生创新训练项目(201610528004)资助
【分类号】：TP301.6;TP334.8
【相似文献】
相关期刊论文 前1条
1 ;Fast parallel algorithms for discrete Gabor expansion and transform based on multirate filtering[J];Science China(Information Sciences);2012年02期
相关博士学位论文 前1条
1 林晨;离散Gabor变换与展开快速并行算法分析与实现[D];安徽大学;2015年
，本文编号：2372882
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2372882.html
