Labeling, Cutting, Grouping: An Efficient Text Line Segmentation Method for Medieval Manuscripts

Abstract

This paper introduces a new way for text-line extraction by integrating deep-learning based pre-classification and state-of-the-art segmentation methods. Text-line extraction in complex handwritten documents poses a significant challenge, even to the most modern computer vision algorithms. Historical manuscripts are a particularly hard class of documents as they present several forms of noise, such as degradation, bleedthrough, interlinear glosses, and elaborated scripts. In this work, we propose a novel method which uses semantic segmentation at pixel level as intermediate task, followed by a text-line extraction step. We measured the performance of our method on a recent dataset of challenging medieval manuscripts and surpassed state-of-the-art results by reducing the error by 80.7%. Furthermore, we demonstrate the effectiveness of our approach on various other datasets written in different scripts. Hence, our contribution is two-fold. First, we demonstrate that semantic pixel segmentation can be used as strong denoising pre-processing step before performing text line extraction. Second, we introduce a novel, simple and robust algorithm that leverages the highquality semantic segmentation to achieve a text-line extraction performance of 99.42% line IU on a challenging dataset.

Combination of deep neural networks and logical rules for record segmentation in historical handwritten registers using few examples

“…The authors demonstrated its genericity by successfully solving five semantic segmentation tasks on historical documents: page extraction, text line extraction, structure detection, decoration detection and photo detection. Albertini et al [3] have used the DeepDIVA framework [2] to obtain high quality semantic segmentation before extracting text-lines. Alaasam et al [1] have used siamese networks at the patch level for semantic segmentation of challenging historical Arabic manuscripts.…”

Layout Analysis and Text Column Segmentation for Historical Vietnamese Steles

“…After semantic segmentation, we further segment the pixels classified as main text into individual text columns using seam carving, a well-established technique for text line segmentation in historical document images [13]. In this work, we use the recently introduced seam carving method proposed by Alberti et al [7], which has achieved a strong performance on several medieval manuscript datasets. The result of text column segmentation are tight polygons around the foreground pixels of the individual text columns.…”

“…Finally, the CCs are clustered with respect to the number of seams to their right in order to form text columns and tight polygons are computed around all main text pixels. For more details on the seam carving method, we refer to [7].…”

“…The latter aims to segment pixels belonging to the main text into separate text columns, which contain Chu Nôm characters written from top to bottom and right to left. For this task, we rely on seam carving for finding the optimal cuts [7]. The resulting text columns can then be used as input for handwriting recognition systems.…”

Separating Chinese Character from Noisy Background Using GAN

“…We can now process a document with both machine-printed text and handwritten text and then recognize them separately [4,5]. Similar applications can be found in the archiving and processing of historical documents [6,7]. In the field of education, related technologies for examination paper autoscoring have emerged, which greatly reduce burden for teachers and students.…”