Ink Enhancement for Ancient Bamboo Manuscripts Using Iterative Restoration-Degradation Adversarial Learning


Chongsheng ZHANG (Henan University, Ludwig-Maximilians-Universität München), Junchao Ma (Henan Univeristy), Wenhao Zhang (Henan Univeristy), Kamel Aouaidjia (Henan Univeristy), Qilong Li (Henan Univeristy), Gaojuan Fan (Henan Univeristy), Christian Heumann (Ludwig-Maximilians-Universität München)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

Bamboo slips are crucial documentary evidences for investigating ancient Chinese history and civilization. However, due to the lengthy deterioration process for about 2000 years, inscriptions on bamboo slips typically suffer from ink fading problem, which seriously impacts their clarity and legibility. This paper proposes BambInk, which is among the first technical initiatives that explore AI-enabled ink enhancement for bamboo slips with faded inks. BambInk is a new self-supervised learning method that devises an iterative restoration-degradation adversarial learning mechanism to progressively enhance the faded inks on the bamboo slips, yet without requiring any annotated data. In specific, BambInk designs two generators which are ink enhancer and ink degrader, and two discriminators for evaluate the effects of the enhanced and degraded images, respectively. Moreover, in the generators, it introduces a dynamic convolution mechanism that integrates features captured via three heterogeneous types of attentions; it also adopts a simple yet effective high-low frequency information differentiation mechanism for extracting the fine-details of the ink traces. Experiments conducted on the real-world bamboo slips dataset demonstrate the effectiveness of our method. Code, data and the enhanced results are available at: https://github.com/cszhangLMU/BambInk.

Citation

@inproceedings{ZHANG_2025_BMVC,
author    = {Chongsheng ZHANG and Junchao Ma and Wenhao Zhang and Kamel Aouaidjia and Qilong Li and Gaojuan Fan and Christian Heumann},
title     = {Ink Enhancement for Ancient Bamboo Manuscripts  Using Iterative Restoration-Degradation Adversarial Learning},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_988/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection