Locating splicing forgery by fully convolutional networks and conditional random field

作者：

Highlights：

•

• Deep neural networks can be used to detect splicing forgery.

• Combining three deep neural networks shows better performance in exposing splicing forgery.

• Outperformed hand-crafted feature based methods.

• Achieved average recall of 98.7% and average accuracy of 73.2% in pixel level in CASIA v2 database.

摘要

•Deep neural networks can be used to detect splicing forgery.•Combining three deep neural networks shows better performance in exposing splicing forgery.•Outperformed hand-crafted feature based methods.•Achieved average recall of 98.7% and average accuracy of 73.2% in pixel level in CASIA v2 database.

论文关键词：Splicing forgery,Deep neural network,Fully convolutional network,Conditional random field

论文评审过程：Received 2 November 2017, Revised 10 April 2018, Accepted 20 April 2018, Available online 30 April 2018, Version of Record 17 June 2018.

论文官网地址：https://doi.org/10.1016/j.image.2018.04.011