@inproceedings{f2f0c0f65c4c42279bca354301c656e4,
title = "Camera source identification with limited labeled training set",
abstract = "This paper investigates the problem of model-based camera source identification with limited labeled training samples. We consider the realistic scenario in which the number of labeled training samples is limited. Ensemble projection (EP) method is proposed by introducing prototype theory into semi-supervised learning. After constructing sub-sets of local binary patterns (LBP) features, several pre-classifiers are established for all labeled and unlabeled samples. According to the ranking of posterior probabilities, several prototype sets are constructed for the ensemble projection. Combining the outputs of all labeled samples from classifiers trained by prototype sets, a new feature vector is generated for camera source identification. Experimental results illustrate that the proposed EP method achieves a notable higher average accuracy than previous algorithms when labeled training samples is limited.",
keywords = "Camera source identification, Ensemble projection, LBP features, Limited labeled training samples",
author = "Yue Tan and Bo Wang and Ming Li and Yanqing Guo and Xiangwei Kong and Yunqing Shi",
note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2016.; 14th International Workshop on Digital-Forensics and Watermarking, IWDW 2015 ; Conference date: 07-10-2015 Through 10-10-2015",
year = "2016",
doi = "10.1007/978-3-319-31960-5_2",
language = "English (US)",
isbn = "9783319319599",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "18--27",
editor = "Isao Echizen and Kim, {Hyoung Joong} and Yun-Qing Shi and Fernando P{\'e}rez-Gonz{\'a}lez",
booktitle = "Digital-Forensics and Watermarking - 14th International Workshop, IWDW 2015, Revised Selected Papers",
address = "Germany",
}