A multistage and multiresolution deep convolutional neural network for inverse halftoning

作者：

Highlights：

• A multistage and multiresolution deep convolutional neural network is proposed for inverse halftoning.

• A multiresolution convolutional neural network is proposed to remove the halftone noise dots completely.

• Residual dense blocks are adopted to generate detail residual maps to further enhance image details.

• Micro-adjust maps are first presented to adjust the restored images globally.

• Proposed method is superior to the state-of-the-art methods.

摘要

•A multistage and multiresolution deep convolutional neural network is proposed for inverse halftoning.•A multiresolution convolutional neural network is proposed to remove the halftone noise dots completely.•Residual dense blocks are adopted to generate detail residual maps to further enhance image details.•Micro-adjust maps are first presented to adjust the restored images globally.•Proposed method is superior to the state-of-the-art methods.

论文关键词：Inverse halftoning,Deep convolutional neural network,Multiresolution neural network,Residual dense block

论文评审过程：Received 24 May 2021, Revised 29 November 2021, Accepted 29 November 2021, Available online 3 December 2021, Version of Record 8 December 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.116358