Even state-of-the-art neural approaches to handwriting recognition struggle when the handwriting is on ruled paper. We thus explore CNN-based methods to remove ruled lines and at the same time retain the parts of the writing overlapping with the ruled line. For that purpose, we devise a method to create a large synthetic dataset for training and evaluation of our models. We show that our best model variants are capable of reconstructing characters that are overlapping with the line to be removed, which is a problem that simpler approaches often fail to solve. On a dataset of children handwriting, we show that removing the ruled lines improves character recognition. We made our synthetic dataset and all experimental code available to foster further research in this area.
CITATION STYLE
Gold, C., & Zesch, T. (2022). CNN-Based Ruled Line Removal in Handwritten Documents. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 13639 LNCS, pp. 530–544). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-21648-0_36
Mendeley helps you to discover research relevant for your work.