Two-stage generative adversarial networks for binarization of color document images

作者：

Highlights：

• We propose a two-stage color document image enhancement and binarization method using generative adversarial networks.

• Four color-independent adversarial networks extract color foreground information from an input image for document image enhancement.

• Two independent adversarial networks with global and local features are trained for image binarization of documents of variable size.

• The proposed method outperforms state-of-the-art algorithms on various datasets.

摘要

•We propose a two-stage color document image enhancement and binarization method using generative adversarial networks.•Four color-independent adversarial networks extract color foreground information from an input image for document image enhancement.•Two independent adversarial networks with global and local features are trained for image binarization of documents of variable size.•The proposed method outperforms state-of-the-art algorithms on various datasets.

论文关键词：Document image binarization,Generative adversarial networks,Optical character recognition,Color document image enhancement

论文评审过程：Received 27 April 2021, Revised 10 May 2022, Accepted 20 May 2022, Available online 23 May 2022, Version of Record 29 May 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108810