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

Abstract

The paper structure of historical prints is sort of a unique fingerprint. Paper with the same origin shows similar chain line distances. As the manual measurement of chain line distances is time consuming, the automatic detection of chain lines is beneficial. We propose an end-to-end trainable deep learning method for segmentation and parameterization of chain lines in transmitted light images of German prints from the 16th Century. We trained a conditional generative adversarial network with a multitask loss for line segmentation and line parameterization. We formulated a fully differentiable pipeline for line coordinates’ estimation that consists of line segmentation, horizontal line alignment, and 2D Fourier filtering of line segments, line region proposals, and differentiable line fitting. We created a dataset of high-resolution transmitted light images of historical prints with manual line coordinate annotations. Our method shows superior qualitative and quantitative chain line detection results with high accuracy and reliability on our historical dataset in comparison to competing methods. Further, we demonstrated that our method achieves a low error of less than 0.7 mm in comparison to manually measured chain line distances.

Extracting chain lines and laid lines from digital images of medieval paper using spectral total variation decomposition

“…They use canny edge detection to additionally extract the watermark. Most recently, both Biendl et al [ 12 ] and Sindel et al [ 13 ] develop deep learning approaches for chain line segmentation and parametrisation from transmitted light images. While [ 12 ] follow a supervised approach to train a UNet for chain line segmentation, [ 13 ] train a conditional generative adversarial network to predict a segmentation mask.…”

“…Most recently, both Biendl et al [ 12 ] and Sindel et al [ 13 ] develop deep learning approaches for chain line segmentation and parametrisation from transmitted light images. While [ 12 ] follow a supervised approach to train a UNet for chain line segmentation, [ 13 ] train a conditional generative adversarial network to predict a segmentation mask. However, both approaches rely on supervised training which in turn requires manually annotated ground truth segmentation masks and is therefore costly.…”

Editorial for Special Issue “Fine Art Pattern Extraction and Recognition”

“…To support art historical research, Sindel et al [12] did not focus on high-level concepts but on the distance measurement of the chain lines in historical prints, which constitute a sort of unique "fingerprint" of their paper structure. Since this process is typically manual, they propose an end-to-end trainable model based on a conditional generative adversarial network that performs line segmentation and parameterization in a multitask fashion.…”