ChainLineNet: Deep-Learning-Based Segmentation and Parameterization of Chain Lines in Historical Prints

Area of Science:

• Art History
• Material Science
• Computer Vision

Background:

• Paper structure, specifically chain line distances, acts as a unique identifier for historical prints.
• Manual measurement of chain line distances is labor-intensive and time-consuming.
• Automating chain line detection is crucial for efficient analysis of historical paper artifacts.

Purpose of the Study:

• To develop an end-to-end trainable deep learning method for automatic segmentation and parameterization of chain lines.
• To enable accurate and reliable identification of paper origins in historical German prints from the 16th Century.
• To provide a faster and more efficient alternative to manual measurement techniques.

Main Methods:

• A conditional generative adversarial network (GAN) was trained using a multitask loss for both line segmentation and parameterization.
• A fully differentiable pipeline was formulated, integrating line segmentation, horizontal line alignment, and 2D Fourier filtering.
• The pipeline includes line region proposals and differentiable line fitting for precise coordinate estimation.
• A dataset of high-resolution transmitted light images with manual line coordinate annotations was created.

Main Results:

• The proposed deep learning method demonstrated superior qualitative and quantitative results in chain line detection.
• High accuracy and reliability were achieved on a dataset of historical German prints.
• The method significantly outperforms competing chain line detection techniques.
• A low error rate of less than 0.7 mm was achieved compared to manual measurements.

Conclusions:

• The developed deep learning approach offers an accurate and efficient solution for analyzing historical print paper structures.
• Automatic chain line detection facilitates the identification of paper origins and enhances historical research.
• This method provides a reliable tool for art historians and material scientists studying historical documents.