A deep learning approach to patch-based image inpainting forensics

作者：

Highlights：

•

• A deep learning approach is proposed to detect patch-based inpainting operation.

• The convolutional encoder–decoder network is developed.

• Inpainting feature learning is supervised by a class label matrix for each image.

• The weighted cross-entropy is designed as the loss function.

摘要

•A deep learning approach is proposed to detect patch-based inpainting operation.•The convolutional encoder–decoder network is developed.•Inpainting feature learning is supervised by a class label matrix for each image.•The weighted cross-entropy is designed as the loss function.

论文关键词：Inpainting,Forensics,Convolutional neural network,Loss function,Compression

论文评审过程：Received 6 November 2017, Revised 11 April 2018, Accepted 28 May 2018, Available online 13 June 2018, Version of Record 19 June 2018.

论文官网地址：https://doi.org/10.1016/j.image.2018.05.015