AADD

Axis-Aligned Document Dewarping

Official implementation of "Axis-Aligned Document Dewarping"

Accepted to AAAI 2026 (Oral Presentation) 🔥

Authors: Chaoyun Wang, I-Chao Shen, Takeo Igarashi, Caigui Jiang

Institutes: Xi'an Jiaotong University, The University of Tokyo

---

📢 News

• 🌟 Our another paper, "Cascaded Robust Rectification for Arbitrary Document Images", has been accepted by IEEE TMM 2026! Check out the ArbDR Project Page for a robust multi-stage rectification framework.
• 📊 We have released the paper results, You can download them via Baidu Cloud (Access Code: j68u).
• 🎉 We are excited to announce that our paper has been accepted to AAAI 2026 as an Oral Presentation!
• [Coming Soon] The full code and models will be released after the final camera-ready submission. Please star ⭐ this repo to stay updated.

🚀 Introduction

Document dewarping is crucial for many applications. However, existing learning-based methods rely heavily on supervised regression with annotated data without fully leveraging the inherent geometric properties of physical documents. Our key insight is that a well-dewarped document is defined by its axis-aligned feature lines. This property aligns with the inherent axis-aligned nature of the discrete grid geometry in planar documents. Harnessing this property, we introduce three synergistic contributions: for the training phase, we propose an axis-aligned geometric constraint to enhance document dewarping; for the inference phase, we propose an axis alignment preprocessing strategy to reduce the dewarping difficulty; and for the evaluation phase, we introduce a new metric, Axis-Aligned Distortion (AAD), that not only incorporates geometric meaning and aligns with human visual perception but also demonstrates greater robustness. As a result, our method achieves state-of-the-art performance on multiple existing benchmarks, improving the AAD metric by 18.2% to 34.5%.

📝 Citation

If you find our work useful, please cite:

@article{wang2025axis,
  title={Axis-Aligned Document Dewarping},
  author={Wang, Chaoyun and Shen, I and Igarashi, Takeo and Jiang, Caigui and others},
  journal={arXiv preprint arXiv:2507.15000},
  year={2025}
}