Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model

Baran Işik; Behçet Uǧur Töreyin

doi:10.1109/MLSP62443.2025.11204243

Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model

Baran Işik, Behçet Uǧur Töreyin

Informatics Institute

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Historical manuscripts suffer from various forms of degradation such as ink bleed, fading, and physical damage, which impairs their readability and usability. This study proposes a novel image restoration framework using Diffusion Probabilistic Models (DPMs) enhanced with OCR-guided finetuning. By integrating a two-stage architecture-comprising a NAFNet-based Initial Predictor and a Diffusion-Based De-noiser-this system restores both low- and high-frequency document details. Evaluations on benchmark datasets demonstrate that the proposed method outperforms traditional techniques in terms of PSNR, SSIM, and OCR accuracy, making it a scalable and robust solution for cultural heritage preservation.

Original language	English
Title of host publication	35th IEEE International Workshop on Machine Learning for Signal Processing
Subtitle of host publication	Signal Processing in the Age of Lorge Language Models, MLSP 2025
Publisher	IEEE Computer Society
ISBN (Electronic)	9798331570293
DOIs	https://doi.org/10.1109/MLSP62443.2025.11204243
Publication status	Published - 2025
Externally published	Yes
Event	35th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2025 - Istanbul, Turkey Duration: 31 Aug 2025 → 3 Sept 2025

Publication series

Name	IEEE International Workshop on Machine Learning for Signal Processing, MLSP
ISSN (Print)	2161-0363
ISSN (Electronic)	2161-0371

Conference

Conference	35th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2025
Country/Territory	Turkey
City	Istanbul
Period	31/08/25 → 3/09/25

Bibliographical note

Publisher Copyright:
© 2025 IEEE.

Keywords

deep learning
diffusion models
document enhancement
Historical manuscripts
image restoration
OCR

Access to Document

10.1109/MLSP62443.2025.11204243

Cite this

Işik, B., & Töreyin, B. U. (2025). Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model. In 35th IEEE International Workshop on Machine Learning for Signal Processing: Signal Processing in the Age of Lorge Language Models, MLSP 2025 (IEEE International Workshop on Machine Learning for Signal Processing, MLSP). IEEE Computer Society. https://doi.org/10.1109/MLSP62443.2025.11204243

Işik, Baran ; Töreyin, Behçet Uǧur. / Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model. 35th IEEE International Workshop on Machine Learning for Signal Processing: Signal Processing in the Age of Lorge Language Models, MLSP 2025. IEEE Computer Society, 2025. (IEEE International Workshop on Machine Learning for Signal Processing, MLSP).

@inproceedings{b25520d4648b48c3bb07cab9cbf9cfe1,

title = "Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model",

abstract = "Historical manuscripts suffer from various forms of degradation such as ink bleed, fading, and physical damage, which impairs their readability and usability. This study proposes a novel image restoration framework using Diffusion Probabilistic Models (DPMs) enhanced with OCR-guided finetuning. By integrating a two-stage architecture-comprising a NAFNet-based Initial Predictor and a Diffusion-Based De-noiser-this system restores both low- and high-frequency document details. Evaluations on benchmark datasets demonstrate that the proposed method outperforms traditional techniques in terms of PSNR, SSIM, and OCR accuracy, making it a scalable and robust solution for cultural heritage preservation.",

keywords = "deep learning, diffusion models, document enhancement, Historical manuscripts, image restoration, OCR",

author = "Baran I{\c s}ik and T{\"o}reyin, \{Beh{\c c}et Uǧur\}",

note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; 35th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2025 ; Conference date: 31-08-2025 Through 03-09-2025",

year = "2025",

doi = "10.1109/MLSP62443.2025.11204243",

language = "English",

series = "IEEE International Workshop on Machine Learning for Signal Processing, MLSP",

publisher = "IEEE Computer Society",

booktitle = "35th IEEE International Workshop on Machine Learning for Signal Processing",

address = "United States",

}

Işik, B & Töreyin, BU 2025, Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model. in 35th IEEE International Workshop on Machine Learning for Signal Processing: Signal Processing in the Age of Lorge Language Models, MLSP 2025. IEEE International Workshop on Machine Learning for Signal Processing, MLSP, IEEE Computer Society, 35th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2025, Istanbul, Turkey, 31/08/25. https://doi.org/10.1109/MLSP62443.2025.11204243

Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model. / Işik, Baran; Töreyin, Behçet Uǧur.
35th IEEE International Workshop on Machine Learning for Signal Processing: Signal Processing in the Age of Lorge Language Models, MLSP 2025. IEEE Computer Society, 2025. (IEEE International Workshop on Machine Learning for Signal Processing, MLSP).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model

AU - Işik, Baran

AU - Töreyin, Behçet Uǧur

PY - 2025

Y1 - 2025

N2 - Historical manuscripts suffer from various forms of degradation such as ink bleed, fading, and physical damage, which impairs their readability and usability. This study proposes a novel image restoration framework using Diffusion Probabilistic Models (DPMs) enhanced with OCR-guided finetuning. By integrating a two-stage architecture-comprising a NAFNet-based Initial Predictor and a Diffusion-Based De-noiser-this system restores both low- and high-frequency document details. Evaluations on benchmark datasets demonstrate that the proposed method outperforms traditional techniques in terms of PSNR, SSIM, and OCR accuracy, making it a scalable and robust solution for cultural heritage preservation.

AB - Historical manuscripts suffer from various forms of degradation such as ink bleed, fading, and physical damage, which impairs their readability and usability. This study proposes a novel image restoration framework using Diffusion Probabilistic Models (DPMs) enhanced with OCR-guided finetuning. By integrating a two-stage architecture-comprising a NAFNet-based Initial Predictor and a Diffusion-Based De-noiser-this system restores both low- and high-frequency document details. Evaluations on benchmark datasets demonstrate that the proposed method outperforms traditional techniques in terms of PSNR, SSIM, and OCR accuracy, making it a scalable and robust solution for cultural heritage preservation.

KW - deep learning

KW - diffusion models

KW - document enhancement

KW - Historical manuscripts

KW - image restoration

KW - OCR

UR - https://www.scopus.com/pages/publications/105022108490

U2 - 10.1109/MLSP62443.2025.11204243

DO - 10.1109/MLSP62443.2025.11204243

M3 - Conference contribution

AN - SCOPUS:105022108490

T3 - IEEE International Workshop on Machine Learning for Signal Processing, MLSP

BT - 35th IEEE International Workshop on Machine Learning for Signal Processing

PB - IEEE Computer Society

T2 - 35th IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2025

Y2 - 31 August 2025 through 3 September 2025

ER -

Işik B, Töreyin BU. Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model. In 35th IEEE International Workshop on Machine Learning for Signal Processing: Signal Processing in the Age of Lorge Language Models, MLSP 2025. IEEE Computer Society. 2025. (IEEE International Workshop on Machine Learning for Signal Processing, MLSP). doi: 10.1109/MLSP62443.2025.11204243

Restoration and Enhancement of Historical Manuscript Images Using Diffusion Model

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this