Rapid image splicing detection based on relevance vector machine

Authors:
Bo Su;Quanqiao Yuan;Yujin Zhang;Mengying Zhai;Shilin Wang
Affiliations:
School of Information Security Engineering, Shanghai Jiao Tong University, China;School of Information Security Engineering, Shanghai Jiao Tong University, China;Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;School of Information Security Engineering, Shanghai Jiao Tong University, China
Venue:
IWDW'12 Proceedings of the 11th international conference on Digital Forensics and Watermaking
Year:
2012

Citing 6
Cited 0

Sparse bayesian learning and the relevance vector machine

The Journal of Machine Learning Research
Exposing digital forgeries by detecting inconsistencies in lighting

MM&Sec '05 Proceedings of the 7th workshop on Multimedia and security
A natural image model approach to splicing detection

Proceedings of the 9th workshop on Multimedia & security
The evidence framework applied to classification networks

Neural Computation
Exposing digital forgeries from JPEG ghosts

IEEE Transactions on Information Forensics and Security
Effective image splicing detection based on image chroma

ICIP'09 Proceedings of the 16th IEEE international conference on Image processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Image splicing detection has become one of the most important topics in the field of information security and much work has been done for that. We focus on its practical application, which considers not only detection rate but also the time consumption. This paper combines Run-length Histogram Features (RLHF) in spatial domain and Markov based features in frequency domain for capturing splicing artifact. Principal Component Analysis (PCA) is adopted to reduce the dimensions of the features in order to reduce the computational complexity in classification. Furthermore, this paper introduces Relevance Vector Machine (RVM) as a classifier and introduces its advantage over Support Vector Machine (SVM) in theory. Simulation shows that the performance of combined features is better than each feature alone. RVM consumes much less test time than SVM at the price of a negligible decline of detection rate. Therefore, the proposed method meets the requirements of a fast and efficient image splicing detection.