Name
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. |
Year | Venue | Field |
|---|---|---|
2015 | IWDW | Training set,Computer vision,Prototype theory,Feature vector,Ranking,Computer science,Local binary patterns,Posterior probability,Artificial intelligence |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
1 | 6 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Yue Tan | 1 | 5 | 2.95 |
| Bo Wang | 2 | 13 | 6.14 |
| Ming Li | 3 | 388 | 37.81 |
| Yanqing Guo | 4 | 39 | 12.24 |
| Xiangwei Kong | 5 | 387 | 37.93 |
| Yun Q. Shi | 6 | 2918 | 199.53 |