TY - GEN
T1 - An Improved Sensor Pattern Noise Estimation Method Based on Edge Guided Weighted Averaging
AU - Zhang, Wen Na
AU - Liu, Yun Xia
AU - Zhou, Jin
AU - Yang, Yang
AU - Law, Ngai Fong
N1 - Funding Information:
Acknowledgments. This work was supported by National Key Research and Development Program (No. 2018YFC0831100), the National Nature Science Foundation of China (No. 61305015, No. 61203269), the National Natural Science Foundation of Shandong Province (No. ZR2017MF057), and Shandong Province Key Research and Development Program, China (No. 2016GGX101022).
Funding Information:
This work was supported by National Key Research and Development Program (No. 2018YFC0831100), the National Nature Science Foundation of China (No. 61305015, No. 61203269), the National Natural Science Foundation of Shandong Province (No. ZR2017MF057), and Shandong Province Key Research and Development Program, China (No. 2016GGX101022).
Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020/11
Y1 - 2020/11
N2 - Sensor Pattern Noise (SPN) has proven to be an effective fingerprint for source camera identification. However, its estimation accuracy is still greatly affected by image contents. In this work, considering the confidence difference in varying image regions, an image edge guided weighted averaging scheme for robust SPN estimation is proposed. Firstly, the edge and non-edge regions are estimated by a Laplacian operator-based detector, based on which different weights are assigned to. Then, the improved SPN estimation is obtained by weighted averaging of image residuals. Finally, an edge guided weighted normalized cross-correlation measurement is proposed as similarity metric in source camera identification (SCI) applications. The effectiveness of the proposed method is verified by SCI experiments conducted on six models from the Dresden data set. Comparison results on different denoising algorithms and varying patch sizes reveal that performance improvement is more prominent for small image patches, which is demanding in real forensic applications.
AB - Sensor Pattern Noise (SPN) has proven to be an effective fingerprint for source camera identification. However, its estimation accuracy is still greatly affected by image contents. In this work, considering the confidence difference in varying image regions, an image edge guided weighted averaging scheme for robust SPN estimation is proposed. Firstly, the edge and non-edge regions are estimated by a Laplacian operator-based detector, based on which different weights are assigned to. Then, the improved SPN estimation is obtained by weighted averaging of image residuals. Finally, an edge guided weighted normalized cross-correlation measurement is proposed as similarity metric in source camera identification (SCI) applications. The effectiveness of the proposed method is verified by SCI experiments conducted on six models from the Dresden data set. Comparison results on different denoising algorithms and varying patch sizes reveal that performance improvement is more prominent for small image patches, which is demanding in real forensic applications.
KW - Edge detection
KW - Sensor pattern noise
KW - Source camera identification
KW - Weighted averaging
UR - http://www.scopus.com/inward/record.url?scp=85097128852&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-62460-6_36
DO - 10.1007/978-3-030-62460-6_36
M3 - Conference article published in proceeding or book
AN - SCOPUS:85097128852
SN - 9783030624590
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 405
EP - 415
BT - Machine Learning for Cyber Security - Third International Conference, ML4CS 2020, Proceedings
A2 - Chen, Xiaofeng
A2 - Yan, Hongyang
A2 - Yan, Qiben
A2 - Zhang, Xiangliang
PB - Springer Science and Business Media Deutschland GmbH
T2 - 3rd International Conference on Machine Learning for Cyber Security, ML4CS 2020
Y2 - 8 October 2020 through 10 October 2020
ER -