TY - JOUR
T1 - An Improved Splicing Localization Method by Fully Convolutional Networks
AU - Chen, Beijing
AU - Qi, Xiaoming
AU - Wang, Yiting
AU - Zheng, Yuhui
AU - Shim, Hiuk Jae
AU - Shi, Yun Qing
N1 - Funding Information:
This work was supported in part by the NSFC under Grants 61572258, 61771231, 61772281, 61602253, and 61672294, in part by the PAPD Fund, and in part by the Qing Lan Project.
PY - 2018
Y1 - 2018
N2 - Liu and Pun proposed a method based on fully convolutional network (FCN) and conditional random field (CRF) to locate spliced regions in synthesized images from different source images. However, their work has two drawbacks: 1) FCN often smooths detailed structures and ignores small objects and 2) CRF is employed as a standalone post-processing step disconnected from the FCN. Therefore, an improved method is proposed in this paper to overcome these two drawbacks. For the first drawback, region proposal network is introduced into the FCN to enhance the learning of object regions. For the second one, the use of CRF is changed to make the whole network an end-to-end learning system. Moreover, the proposed method uses three FCNs (FCN8, FCN16, and FCN32) with different upsampling layers, and all the three FCNs are initialized from VGG-16 network. Experimental results on three publicly available datasets (DVMM dataset, CASIA v1.0 dataset, and CASIA v2.0 dataset) demonstrate that the proposed method can achieve a better performance than the state-of-the-art methods including some conventional methods and some deep learning-based methods.
AB - Liu and Pun proposed a method based on fully convolutional network (FCN) and conditional random field (CRF) to locate spliced regions in synthesized images from different source images. However, their work has two drawbacks: 1) FCN often smooths detailed structures and ignores small objects and 2) CRF is employed as a standalone post-processing step disconnected from the FCN. Therefore, an improved method is proposed in this paper to overcome these two drawbacks. For the first drawback, region proposal network is introduced into the FCN to enhance the learning of object regions. For the second one, the use of CRF is changed to make the whole network an end-to-end learning system. Moreover, the proposed method uses three FCNs (FCN8, FCN16, and FCN32) with different upsampling layers, and all the three FCNs are initialized from VGG-16 network. Experimental results on three publicly available datasets (DVMM dataset, CASIA v1.0 dataset, and CASIA v2.0 dataset) demonstrate that the proposed method can achieve a better performance than the state-of-the-art methods including some conventional methods and some deep learning-based methods.
KW - Splicing localization
KW - conditional random field
KW - fully convolutional network
KW - region proposal network
UR - http://www.scopus.com/inward/record.url?scp=85056496731&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85056496731&partnerID=8YFLogxK
U2 - 10.1109/ACCESS.2018.2880433
DO - 10.1109/ACCESS.2018.2880433
M3 - Article
AN - SCOPUS:85056496731
SN - 2169-3536
VL - 6
SP - 69472
EP - 69480
JO - IEEE Access
JF - IEEE Access
M1 - 8531703
ER -