Transfer subspace learning based on structure preservation for JPEG image mismatched steganalysis

作者：

Highlights：

• A novel transfer learning method is proposed for JPEG image mismatched steganalysis.

• A Frobenius-norm regularization is adopted to preserve the structure of features.

• A structured sparsity constraint is used to reduce the impact of noises and outliers.

• The validity of the proposed method is verified by extensive comparison experiments.

摘要

•A novel transfer learning method is proposed for JPEG image mismatched steganalysis.•A Frobenius-norm regularization is adopted to preserve the structure of features.•A structured sparsity constraint is used to reduce the impact of noises and outliers.•The validity of the proposed method is verified by extensive comparison experiments.

论文关键词：Mismatch,Steganalysis,JPEG image,Transfer subspace learning,Structure preservation

论文评审过程：Received 11 September 2019, Revised 3 May 2020, Accepted 25 October 2020, Available online 31 October 2020, Version of Record 4 November 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.116052