Text Line Segmentation for Challenging Handwritten Document Images using Fully Convolutional Network

Barakat, Berat Kurar; Droby, Ahmad; Kassis, Majeed; El-Sana, Jihad

doi:10.1109/icfhr-2018.2018.00072

Cited by 43 publications

(24 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It was performed using a deep FCN based on the dilated convolutions. In [2], Barakat et al present a line segmentation method for historical document images. They have estimated a line mask using a FCN.…”

Section: Related Workmentioning

confidence: 99%

“…Moreover, the text lines in a handwritten document image is more challenging than a printed document image having a flat-bed surface. The text line segmentation techniques developed till now are either language-specific like for Kannada [1], English [30], Hindi [14], Gujrati [6], Bangla [20], Arabic [2], or script independent [3,29]. These methods are tested on line segmentation on a flat-bed surface.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, it is reported that the learning-based approaches, like [22]work with greater accuracy in a similar domain. Although some methods present in [2,23] uses a learning-based approach to segment text-lines, the use of deep artificial neural networks are to be explored yet in the domain of text line segmentation.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Text Line Segmentation: A FCN Based Approach

Minj

Garai

Mandal

2021

Communications in Computer and Information Science

View full text Add to dashboard Cite

Text line segmentation is a prerequisite for most of the document processing systems. However, for handwritten/warped documents, it is not straightforward to segment the text lines. This work proposes a learning-based text line segmentation method from document images. This work can tackle complex layouts present in a camera captured or handwritten document images along with printed flat-bed scanned English documents. The method also works for Alphasyllabrary scripts like Bangla. Segmentation of Bangla handwritten text is quite challenging because of its unique characteristics. The proposed approach of line segmentation relies on fully convolutional networks (FCNs). To improve the performance of the method, we introduce a post-processing step. The model is trained and tested on our dataset along with the cBAD dataset. We develop the model in such a way that it can be trained and tested in a machine that has limited access to highly computational accessories like GPU. The results of our experiments are encouraging.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Text Line Segmentation: A FCN Based Approach

Minj

Garai

Mandal

2021

Communications in Computer and Information Science

View full text Add to dashboard Cite

show abstract

“…2. [12,15], the object detection based HTLS [16], and the start-and-follow based HTLS [20]. For fair comparisons, we always use the default settings of a solution provided in its public git repository, including but are not limited to pre-and post-processing, learning rate, and optimizer.…”

Section: Ablation Study On Monotone Enforcementmentioning

confidence: 99%

Linecounter: Learning Handwritten Text Line Segmentation By Counting

Deng¹,

Zhou³

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Handwritten Text Line Segmentation (HTLS) is a low-level but important task for many higher-level document processing tasks like handwritten text recognition. It is often formulated in terms of semantic segmentation or object detection in deep learning. However, both formulations have serious shortcomings. The former requires heavy postprocessing of splitting/merging adjacent segments, while the latter may fail on dense or curved texts. In this paper, we propose a novel Line Counting formulation for HTLS -that involves counting the number of text lines from the top at every pixel location. This formulation helps learn an end-to-end HTLS solution that directly predicts per-pixel line number for a given document image. Furthermore, we propose a deep neural network (DNN) model LineCounter to perform HTLS through the Line Counting formulation. Our extensive experiments on the three public datasets (ICDAR2013-HSC [1], HIT-MW [2], and VML-AHTE [3]) demonstrate that LineCounter outperforms state-of-the-art HTLS approaches.

show abstract

“…Chen et al [25] explore the use of Fully Convolutional Networks (FCN) for the same datasets. Barakat et al [26] propose a FCN for segmenting closely spaced, arbitrarily oriented text lines from an Arabic manuscript dataset. The mentioned approaches, coupled with efforts to conduct competitions on various aspects of historical document layout analysis have aided progress in this area [27]- [29].…”

Section: Related Workmentioning

confidence: 99%

Indiscapes: Instance Segmentation Networks for Layout Parsing of Historical Indic Manuscripts

Prusty

Aitha

Trivedi

et al. 2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Historical palm-leaf manuscript and early paper documents from Indian subcontinent form an important part of the world's literary and cultural heritage. Despite their importance, large-scale annotated Indic manuscript image datasets do not exist. To address this deficiency, we introduce Indiscapes, the first ever dataset with multi-regional layout annotations for historical Indic manuscripts. To address the challenge of large diversity in scripts and presence of dense, irregular layout elements (e.g. text lines, pictures, multiple documents per image), we adapt a Fully Convolutional Deep Neural Network architecture for fully automatic, instance-level spatial layout parsing of manuscript images. We demonstrate the effectiveness of proposed architecture on images from the Indiscapes dataset. For annotation flexibility and keeping the non-technical nature of domain experts in mind, we also contribute a custom, webbased GUI annotation tool and a dashboard-style analytics portal. Overall, our contributions set the stage for enabling downstream applications such as OCR and word-spotting in historical Indic manuscripts at scale.

show abstract

Text Line Segmentation for Challenging Handwritten Document Images using Fully Convolutional Network

Cited by 43 publications

References 17 publications

Text Line Segmentation: A FCN Based Approach

Text Line Segmentation: A FCN Based Approach

Linecounter: Learning Handwritten Text Line Segmentation By Counting

Indiscapes: Instance Segmentation Networks for Layout Parsing of Historical Indic Manuscripts

Contact Info

Product

Resources

About