王波
Professor Supervisor of Doctorate Candidates Supervisor of Master's Candidates
Main positions:Professor
Gender:Male
Alma Mater:Dalian University of Technology
Degree:Doctoral Degree
School/Department:School of Information and Communication Engineering
Discipline:Signal and Information Processing
Business Address:A512, Haishan Building
Contact Information:bowang@dlut.edu.cn
E-Mail:bowang@dlut.edu.cn
Hits:
Indexed by:会议论文
Date of Publication:2015-03-28
Page Number:448-454
Key Words:数字图像;相机来源;鉴定方法;有限样本;半监督学习
Abstract:针对已有的数字图像相机来源鉴别取证算法,尽管在实验室环境下能够取得较高的准确率,但是其基于统计特征和多类分类的模式,决定了只能在拥有大量有标签训练样本的条件下达到较好的检测性能.针对有限样本条件下的图像相机来源取证鉴別问题,论文利用局部二进制模型特征(LBP)、图像质量特征(IQM)、色彩滤波阵列插值特征(CFA)三类统计特征集作为来源鉴别特征集合,通过自学习方法和协同训练方法,将半监督学习引入相机来源鉴别中,提出并实现了在有限样本条件下数字图像来源鉴别的有效取证方法.实验结果表明基于半监督学习的来源鉴别方法,能够在标签样本仅有10个的情况下,达到90％以上的平均鉴别准确率.
I am not a star professor, but I am working on the road to a star professor of tomorrow. For this reason, self-motivated graduate students are desired to my research group. Self-motivated attitude and initiative are the most important characteristics in our laboratory. Students with preliminary knowledge on signal (image/video) processing and programming skills are most welcome. For your academic trip in my group, I will devote myself to training your FIVE abilities: Intellectual skills, Communication skills, Personality characteristics, Habit of work, Mechanical skills, which are considered as the most important abilities of a graduated student.
http://ice.dlut.edu.cn/WangBo/index.html
