Search

We present an efficient and effective approach to train OCR engines using the Aletheia document analysis system. All components required for training are seamlessly integrated into Aletheia: training data preparation, the OCR engine's training processes themselves, text recognition, and quantitative evaluation of the trained engine. Such a comprehensive training and evaluation system, guided through a GUI, allows for iterative incremental training to achieve best results. The widely used Tesseract OCR engine is used as a case study to demonstrate the efficiency and effectiveness of the proposed approach. Experimental results are presented validating the training approach with two different historical datasets, representative of recent significant digitisation projects. The impact of different training strategies and training data requirements is presented in detail. [ABSTRACT FROM AUTHOR]

Due to storage conditions and material's non‐planar shape, geometric distortion of the two‐dimensional content is widely present in scanned document images. Effective geometric restoration of these distorted document images considerably increases character recognition rate in large‐scale digitisation. For large‐scale digitisation of historical books, geometric restoration solutions expect to be accurate, generic, robust, unsupervised and reversible. However, most methods in the literature concentrate on improving restoration accuracy for specific distortion effect, but not their applicability in large‐scale digitisation. This study proposes an effective mesh based geometric restoration system (GRLSD) for large‐scale distorted historical document digitisation. In this system, an automatic mesh generation based dewarping tool is proposed to geometrically model and correct arbitrary warping historical documents. An XML‐based mesh recorder is proposed to record the mesh of distortion information for reversible use. A graphic user interface (GUI) toolkit is designed to visually display and manually manipulate the mesh for improving geometric restoration accuracy. Experimental results show that the proposed automatic dewarping approach efficiently corrects arbitrarily warped historical documents, with an improved performance over several state‐of‐the‐art geometric restoration methods. By using XML mesh recorder and GUI toolkit, the GRLSD system greatly aids users to flexibly monitor and correct ambiguous points of mesh for the prevention of damaging historical document images without distortions in large‐scale digitalisation. [ABSTRACT FROM AUTHOR]

This article discusses two major initiatives tasked with developing tools to improve optical character recognition (OCR) or the mechanical keying of texts that are digitally available only as page images. The two initiatives are the IMProving ACcess to Text Project in Europe and the Early Modern OCR Project in the USA. Because of dealing with a multilayered problem like OCR technologies and having to collaborate with radically interdisciplinary and international team members, the two projects developed techniques that we call Agile Project Management, outlined in this essay with rationales for their use. [ABSTRACT FROM AUTHOR]

A preface to "International Journal on Document Analysis and Recognition" is presented.

With the evolution of electronic devices, such as 3D cameras, addressing the challenges of text localization in 3D video (e.g., for indexing) is increasingly drawing the attention of the multimedia and video processing community. Existing methods focus on 2D video and their performance in the presence of the challenges in 3D video, such as shadow areas associated with text and irregularly sized and shaped text, degrades. This paper proposes the first approach that successfully addresses the challenges of 3D video in addition to those of 2D. It employs a number of innovations, among which, the first is the Generalized Gradient Vector Flow (GGVF) for dominant points detection. The second is the Wavefront concept for text candidate point detection from those dominant points. In addition, an Adaptive B-Spline Polygon Curve Network (ABS-Net) is proposed for accurate text localization in 3D videos by constructing tight fitting bounding polygons using text candidate points. Extensive experiments on custom (3D video) and standard datasets (2D video and scene text) show that the proposed method is practical and useful, and overall outperforms existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Abstract: In a variety of documents, ranging from forms to archive documents and books with annotations, machine printed and handwritten text may coexist in the same document image, raising significant issues within the recognition pipeline. It is, therefore, necessary to separate the two types of text so that it becomes feasible to apply different recognition methodologies to each modality. In this paper, a new approach is proposed which strives towards identifying and separating handwritten from machine printed text using the Bag of Visual Words model (BoVW). Initially, blocks of interest are detected in the document image. For each block, a descriptor is calculated based on the BoVW. The final characterization of the blocks as Handwritten, Machine Printed or Noise is made by a decision scheme which relies upon the combination of binary SVM classifiers. The promising performance of the proposed approach is shown by using a consistent evaluation methodology which couples meaningful measures along with new datasets dedicated to the problem upon consideration. [Copyright &y& Elsevier]

In the case of freestyle everyday handwritten documents, writing, erasing, striking out, and overwriting are common behaviors of the writers. This not cleanly-written text poses significant challenges for text line segmentation. Accurate text line segmentation in handwritten documents is essential to the success of several real-world applications, such as answer script evaluation, fraud document identification, writer identification, document age estimation and writer gender classification, to name a few. This paper proposes the first, to the authors' best knowledge, text line segmentation approach that is applicable in the presence of both cleanly-written and struck-out text. The approach consists of three steps. In the first step, components - at the word level - are detected in the input handwritten document images (containing both cleanly-written and struck-out text) based on stroke width information estimation, filtering of noise, and morphological operations. In the second step, the struck-out components are identified using the DenseNet deep learning model and treated differently to clean text in further analysis. In the third step, geometrical spatial features, the direction between candidate components and the overall text line, and the common overlapping region between adjacent components are evaluated to progressively form text lines. To evaluate the proposed steps and compare the proposed method to the state-of-the-art, experiments have been conducted on a new problem-focused dataset containing instances of struck-out text in handwritten documents, as well as on two standard datasets (ICDAR2013 text line segmentation contest dataset and ICDAR2019 HDRC dataset) to show the proposed steps are effective and useful, with superior performance compared to existing methods. [ABSTRACT FROM AUTHOR]