Complete System for Text Line Extraction Using Convolutional Neural Networks and Watershed Transform

Pastor-Pellicer, Joan; Afzal, Muhammad Zeshan; Liwicki, Marcus; Castro-Bleda, María José

doi:10.1109/das.2016.58

Cited by 36 publications

(18 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The closest work related to our method are multi-step methods, presented by Pastor et al [14] and Gruüning et al [15]. The former employs a multi-stage deep learning approach to detect text regions followed by watershed-transform as post-processing step.…”

Section: Related Workmentioning

confidence: 99%

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

Alberti

Vögtlin

Pondenkandath

et al. 2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

Self Cite

View full text Add to dashboard Cite

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.

show abstract

Section: Related Workmentioning

confidence: 99%

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

Alberti

Vögtlin

Pondenkandath

et al. 2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

Self Cite

View full text Add to dashboard Cite

show abstract

“…More recently, [12] used CNNs for textline extraction also aiming at a robust method explicitly focusing on historical documents. However, textlines are segmented at a different level of detail, returning a surrounding polygon (Main Body Area) at pixel level.…”

Section: Introductionmentioning

confidence: 99%

Baseline Detection in Historical Documents Using Convolutional U-Nets

Fink¹,

Layer²,

Mackenbrock³

et al. 2018

2018 13th IAPR International Workshop on Document Analysis Systems (DAS)

View full text Add to dashboard Cite

Baseline detection is still a challenging task for heterogeneous collections of historical documents. We present a novel approach to baseline extraction in such settings, turning out the winning entry to the ICDAR 2017 Competition on Baseline detection (cBAD). It utilizes deep convolutional nets (CNNs) for both, the actual extraction of baselines, as well as for a simple form of layout analysis in a pre-processing step. To the best of our knowledge it is the first CNN-based system for baseline extraction applying a U-net architecture and sliding window detection, profiting from a high local accuracy of the candidate lines extracted. Final baseline post-processing complements our approach, compensating for inaccuracies mainly due to missing context information during sliding window detection. We experimentally evaluate the components of our system individually on the cBAD dataset. Moreover, we investigate how it generalizes to different data by means of the dataset used for the baseline extraction task of the ICDAR 2017 Competition on Layout Analysis for Challenging Medieval Manuscripts (HisDoc). A comparison with the results reported for HisDoc shows that it also outperforms the contestants of the latter.

show abstract

“…Then, the text lines are extracted by superimposing the components with text line pattern mask. Another method is based on convolution neural network with watershed transform, which is proposed in [10] to estimate the text area between the baseline and the corpus line.…”

Section: Introductionmentioning

confidence: 99%

Segmentation of Arabic Handwritten Documents into Text Lines using Watershed Transform

Souhar¹,

Boulid²,

Ameur³

et al. 2017

IJIMAI

View full text Add to dashboard Cite

A crucial task in character recognition systems is the segmentation of the document into text lines and especially if it is handwritten. When dealing with non-Latin document such as Arabic, the challenge becomes greater since in addition to the variability of writing, the presence of diacritical points and the high number of ascender and descender characters complicates more the process of the segmentation. To remedy with this complexity and even to make this difficulty an advantage since the focus is on the Arabic language which is semi-cursive in nature, a method based on the Watershed Transform technique is proposed. Tested on «Handwritten Arabic Proximity Datasets» [21] a segmentation rate of 93% for a 95% of matching score is achieved.

show abstract

Complete System for Text Line Extraction Using Convolutional Neural Networks and Watershed Transform

Cited by 36 publications

References 21 publications

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

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

Baseline Detection in Historical Documents Using Convolutional U-Nets

Segmentation of Arabic Handwritten Documents into Text Lines using Watershed Transform

Contact Info

Product

Resources

About