Related Experiment Video
Updated: Jun 13, 2026
09:16
High-resolution Patterning Using Two Modes of Electrohydrodynamic Jet: Drop on Demand and Near-field Electrospinning
Published on: July 10, 2018
User-assisted ink-bleed reduction.
Yi Huang1, Michael S Brown, Dong Xu
1School of Computer Engineering, Nanyang Technological University, Singapore. hu0005yi@ntu.edu.sg
Summary
This study introduces a user-assisted method to reduce ink-bleed in old manuscripts. The approach uses user examples to train a dual-layer Markov random field (MRF) for accurate pixel labeling and improved manuscript restoration.
Area of Science:
- Digital image processing
- Manuscript restoration
- Computer vision
Background:
- Old manuscripts suffer from ink-bleed, obscuring text and historical data.
- Existing ink-bleed reduction algorithms often require extensive parameter tuning or make assumptions about ink properties.
Purpose of the Study:
- To develop a novel user-assisted approach for effective ink-bleed reduction in historical manuscripts.
- To improve the clarity and readability of degraded manuscript content.
Main Methods:
- User provides examples of foreground ink, ink-bleed, and background.
- Likelihoods computed from user data serve as data costs for a dual-layer Markov random field (MRF).
- Simultaneous labeling of pixels on both manuscript sides using the MRF model.
Main Results:
- Achieves superior ink-bleed reduction compared to existing methods.
- Requires minimal user input and no prior assumptions on ink-bleed intensity.
- Demonstrates robustness across different manuscript degradation levels.
Conclusions:
- The user-assisted MRF approach offers an efficient and effective solution for manuscript ink-bleed restoration.
- This method enhances accessibility to historical documents by improving text legibility.
- The framework provides a flexible and adaptable tool for digital humanities and archival science.
