Document Recognition Research Articles

The Usage of Sources in recognition of original Historical Documents

نسخه‌های اصل، یکی از موثق‌ترین و مستندترین منابع اولیه در پژوهش‌های تاریخی است.هرگونه خدشه در اصالت این منابع، مسیر پژوهش‌ها را به انحراف می‌کشاند.احراز اعتبار، و تشخیص اولیه یا ثانویه بودن این منابع، نخستین شرط تحقیقات پژوهشی است.دست‌یابی به روش‌های شناسایی اعتبار نسخه‌های اصل وابسته به تاریخ، هدف این نوشتار است.بهره‌برداری از اطلاعات مآخذ(منابع و مطالعات)، به عنوان یکی از روش‌های تشخیص اصالت نسخه‌های اصل، در این تحقیق، مورد بررسی قرار می‌گیرد.این پژوهش، از طریق نقد بیرونی یا سندی نسخه‌های مورد بررسی درصدد تبیین موضوع است.این روش با توصیف، مقایسه و تحلیل محقق می‌شود.نتیجه مطالعه بر روی نسخه‌های مورد بررسی در این پژوهش نشان می‌دهد، الگوی ارائه شده، یعنی الگوی استفاده از اطلاعات مآخذ، در تعیین اعتبار نسخه‌های اصل غیرتاریخی نیز کاربرد دارد.بهره‌گیری از اطلاعات مآخذ در شناسایی برخی نسخه‌های اصل، شرطی لازم به شمار می‌آید؛ اما شرطی کافی نیست.تعیین اعتبار نسخه‌های اصل، نیازمند توجه محقق، به مجموعه‌ای از عوامل و روش‌های تشخیصی است.روش‌شناسی نسخ اصل را تشکیل می‌دهد.روش‌هایی که با وارسی دقیق و مطالعه نسخه‌ها در جستجوی قراین گوناگون به دست می‌آید.

Recognition and retrieval of mathematical expressions

Document recognition and retrieval technologies complement one another, providing improved access to increasingly large document collections. While recognition and retrieval of textual information is fairly mature, with wide-spread availability of optical character recognition and text-based search engines, recognition and retrieval of graphics such as images, figures, tables, diagrams, and mathematical expressions are in comparatively early stages of research. This paper surveys the state of the art in recognition and retrieval of mathematical expressions, organized around four key problems in math retrieval (query construction, normalization, indexing, and relevance feedback), and four key problems in math recognition (detecting expressions, detecting and classifying symbols, analyzing symbol layout, and constructing a representation of meaning). Of special interest is the machine learning problem of jointly optimizing the component algorithms in a math recognition system, and developing effective indexing, retrieval and relevance feedback algorithms for math retrieval. Another important open problem is developing user interfaces that seamlessly integrate recognition and retrieval. Activity in these important research areas is increasing, in part because math notation provides an excellent domain for studying problems common to many document and graphics recognition and retrieval applications, and also because mature applications will likely provide substantial benefits for education, research, and mathematical literacy.

A general approach for multi-oriented text line extraction of handwritten documents

The multi-orientation occurs frequently in ancient handwritten documents, where the writers try to update a document by adding some annotations in the margins. Due to the margin narrowness, this gives rise to lines in different directions and orientations. Document recognition needs to find the lines everywhere they are written whatever their orientation. This is why we propose in this paper a new approach allowing us to extract the multi-oriented lines in scanned documents. Because of the multi-orientation of lines and their dispersion in the page, we use an image meshing allowing us to progressively and locally determine the lines. Once the meshing is established, the orientation is determined using the Wigner–Ville distribution on the projection histogram profile. This local orientation is then enlarged to limit the orientation in the neighborhood. Afterward, the text lines are extracted locally in each zone basing on the follow-up of the orientation lines and the proximity of connected components. Finally, the connected components that overlap and touch in adjacent lines are separated. The morphology analysis of the terminal letters of Arabic words is here considered. The proposed approach has been experimented on 100 documents reaching an accuracy of about 98.6%.

Agricultural Trade Facilitation in Asia: Prioritising the Invisible Infrastructure

In the backdrop of the growing agricultural and food trade across Asia, this paper argues that the installed food safety and associated infrastructure could be seen as only necessary conditions whereas the sufficient conditions would be to make an enabling policy environment which would reduce the overall transaction costs of trade. Imports also assume equal importance and the arrival of duty-free raw materials for further processing and value addition enables more employment generation, higher income for stakeholders, and forward movement in the value chain. As tariffs do not account for a substantial influence on the course of trade and prices of many farm commodities, the attention has to turn towards the enabling policy regime specific to commodities and thereby the development of infrastructure which would encourage value addition and re-exports. Trade facilitation and infrastructure are often taken in the general sense at the policy level and only partly address the specific issues related to the reduction of risks and transaction costs in the context of agricultural trade, i.e. the invisible infrastructure such as easy documentation, customs procedures, regulatory regimes and mutual recognition. Different commodities have different requirements in terms of costs, time and reliability of logistics and trade facilitating policies are to be formulated taking into consideration a wider set of factors and indicators in the context of agricultural trade. JEL Codes: F13, F14, F18, Q17, Q18.

Special issue on document recognition and retrieval 2009

This special issue brings together papers presented at the 16th Document Recognition and Retrieval Conference (DRR) held in January 2009 in San Jose, CA, as part of the IS&T/ SPIE Electronic Imaging Symposium. The DRR conference aims at presenting approaches which may improve the extraction of information from various types of documents: scanned, web-based, online or camera-based documents. These approaches may concern the early stages of information extraction such as image enhancement, binarization, or document structure extraction. They may also concern higher stages of information extraction such as OCR, handwriting recognition, or symbol recognition. When these stages are achieved or when an electronic version is already available, information extraction and retrieval techniques can be applied. Related studies include document or writer/signature authentication. This special issue combines seven papers that discuss the different problems encountered in document recognition and retrieval, ranging from geometric distortion detection to handwriting recognition and writer identification. Thework by vanBeusekom, Shafait andBreuel deals with problems arising when a large number of images are scanned (fax images or images received in advanced mailrooms) with a variety of scan orientations. The proposedmethod detects in a combined schemewhether a scannedpagehas been scanned in the correct orientation and the skew angle.

Document Recognition for a Million Books

As initiatives such as Google Book Search (http://books.google.com/) and the Open Content Alliance (http://www.opencontentalliance.org/) advance efforts to digitize millions of books, there is great potential to make available vast amounts of information. To truly unlock this knowledge, however, it will be necessary to process the resulting digital page images to recognize important content, including both the semantic and structural aspects. Given the vast diversity of fonts, symbols, tables, languages and a host of other elements, it will be necessary to create flexible, modular, scalable document recognition systems. Document recognition involves extracting features from the images and even transcriptions of other documents in order to group diverse content.

A non-contact method of capturing low-resolution text for OCR

Document recognition is a lively research area with much effort concentrated on optical character recognition. Less attention is paid to locating and extracting text from the general (non-desktop, non-scanner) environment. Such contact-free extraction of text from a general scene has applications in the context of wearable computing, robotic vision, point and click document capture, or as an aid for visually handicapped people. Here, a novel automatic text reading system is introduced using an active camera focused on text regions already located in the scene (using our recent work). Initially, a located region of text is analysed to determine the optimal zoom that would foveate onto it. Then a number of images are captured over the text region to construct a high-resolution mosaic composite of the whole region. This magnified image of the text is suitable for reading by humans or for recognition by OCR, or even for text-to speech synthesis. Although we employed a low resolution camera, we still obtained very good results.

The architecture of TrueViz: a groundTRUth/metadata editing and VIsualiZing ToolKit

Tools for visualizing and creating groundtruth and metadata are crucial for document image analysis research. In this paper we describe TrueViz (TRUEVIZ User's Manual, August 2000; Proceedings of the SPIE Conference on Document Recognition and Retrieval, San Jose, CA, 2001, pp. 1–12), which is a tool for visualizing and editing groundtruth/metadata. We first describe the groundtruthing task and the requirements for any interactive groundtruthing tool. Next we describe the system design of TrueViz and discuss how a user can use it to create groundtruth. TrueViz is implemented in the Java programming language and works on various platforms including Windows and Unix. TrueViz reads and stores groundtruth/metadata in XML format, and reads a corresponding image stored in TIFF image file format. Multilingual text editing, display, and search modules based on the Unicode representation for text are also provided. This software is being made available free of charge to researchers.

Information retrieval and OCR

IR and OCR have largely developed independent standards and metrics, with OCR focused on literal accuracy, and IR focused on essential "content/meaning". With more and more media not only paper, but in multiple image formats, the opportunities and challenges for OCR on new formats -- video and still images -- are enormous. While OCR is assessed in metrics that emphasize words and characters, IR has learned to apply end-to-end metrics that ask whether the needs of the users can be met by existing systems. The same considerations apply also to the problem of providing permanent worldwide access to millions of pages of legacy print documents, representing the shared human record as it existed until just a few years ago.The International Society for Optical Engineering (SPIE) has held a series of Document Recognition and Retrieval (DRR) conferences. The tenth, DRR X will be held in January 2003, in Santa Clara California. In 2001, Dan LoPresti of Bell Labs decided that the area would benefit from more intense collaboration between those who specialize in finding the words on a page image, and those researchers who know how to find the right documents, given the words. He invited Paul Kantor (Rutgers) to join the DRR Chairs, and together they invited Dave Lewis (Consultant) to give a keynote address at DRR VIII. Dan then stepped down. Paul chaired DRR IX (2002) and then handed the reins to Tapas Kanungo (IBM, Almaden) and together they invited Jamie Callan (CMU), David Grossman (IIT) and Alex Hauptmann (CMU) to join the conference committee for DRR X.To improve communication between SIGIR and DRR, this group proposed a SIGIR workshop on this area. The workshop on "Information Retrieval and OCR: From Converting Content to Grasping Meaning" was intended to stimulate cross-fertilization between OCR and IR, in hopes that better use of IR will enable the OCR community to avoid expensive hand processing, and to demonstrate that the combination of present static and dynamic image processing and present state-of-the-art robust information retrieval can generate substantial advances in both extraction of messages from image streams and conversion of existing paper variants. It solicited papers dealing with future applications, such as the indexing and retrieval of text embedded in static or video graphic images, with problems of skew, distortion, and obscuration, as well as state-of-the-art discussions of the storage and retrieval of handwritten or print legacy materials.The workshop was held on August 15, 2002 in Tampere, Finland, immediately following the SIGIR 2002 conference. Although the workshop was intended to appeal to a wide range of IR and OCR researchers (and indeed was proposed at the request of colleagues from the OCR community), it primarily drew people with a background in IR. About a dozen people participated. The small size allowed a very interactive, seminar-style format and very vigorous discussion between and during presentations. Most presentations ran 30% to 50% longer than planned, and our impression is that most of the participants found it very productive.

Qualitative representation of positional information

A framework for the qualitative representation of positional information in a two-dimensional space is presented. Qualitative representations use discrete quantity spaces, where a particular distinction is introduced only if it is relevant to the context being modeled. This allows us to build a flexible framework that accommodates various levels of granularity and scales of reasoning. Knowledge about position in large-scale space is commonly represented by a combination of orientation and distance relations, which we express in a particular frame of reference between a primary object and a reference object. While the representation of orientation comes out to be more straightforward, the model for distances requires that qualitative distance symbols be mapped to geometric intervals in order to be compared; this is done by defining structure relations that are able to handle, among others, order of magnitude relations; the frame of reference with its three components (distance system, scale, and type) captures the inherent context dependency of qualitative distances. The principal aim of the qualitative representation is to perform spatial reasoning: as a basic inference technique, algorithms for the composition of positional relations are developed with respect to same and different frames of reference. The model presented in this paper has potential applications in areas as diverse as Geographical Information Systems (GIS), Computer Aided Design (CAD), and Document Recognition.

Postal address reading in real time

In 1994, the U.S. Postal Service (USPS) delivered 177 billion pieces of mail to 125 million delivery addresses. To assist in processing this large volume of mail, USPS and the Center for Document Analysis and Recognition (CEDAR) have been exploring automation methods for processing letter mail with handwritten addresses. The CEDAR system for locating and reading destination address blocks on letter mail pieces operates at real-time rates. The system architecture uses a variety of pipelining and multiprocessor techniques to achieve the required processing speed and minimize latency. Image processing operations are performed using specialized hardware operating at 103 ns/pixel. This hardware includes CCITT Group 3 (FAX) decompression, image enhancement, and conversion of the image representation from pixel-based representation to symbolic representation. This symbolic representation of the image can be processed in real time using software running on general-purpose processors. This software performs the address block location (ABL), handwritten address interpretation (HWAI), and data base querying needed to determine the delivery point of the mail piece. The current system can correctly locate and produce a five-digit ZIP code with an accuracy of 66% on a test set of 1566 handwritten mail pieces. A unique delivery point (e.g., ZIP + 4 code) can be determined on these same mail pieces 26% of the time. © 1996 John Wiley & Sons, Inc.

A system to read names and addresses on tax forms

The reading of names and addresses is one of the most complex tasks in automated forms processing. This paper describes an integrated real-time system to read names and addresses on tax forms of the U.S. Internal Revenue Service. The Name and Address Block Reader (NABR) system accepts both machine-printed and hand-printed address block images as input. The application software has two major steps: document analysis (connected component analysis, address block extraction, label detection, hand-print/machine-print discrimination) and document recognition. Document recognition has two nonidentical streams for machine-print and hand-print: the key steps are address parsing, character recognition, word recognition, and postal database lookup. (ZIP+4 and City-State-ZIP files.) System output is a packet containing the results of recognition together with database access status file. Real-time throughput (8500 forms/h) is achieved by employing a loosely coupled multiprocessing architecture where successive input images are distributed to available address recognition processors. The functional architecture, software design, system architecture, and the hardware implementation are described. Performance evaluation on machine-printed and handwritten addresses are presented.

A document recognition system and its applications

This paper describes a document entry system called the Document Recognition System (DRS), which facilitates the conversion of printed documents into electronic form. DRS was developed on a personal computer (PC) with an adapter card for recognizing more than 3000 Kanji characters. It provides a flexible framework for object-oriented management of data and processing modules. The framework allows the user to change the combination of processing modules and to select pipelining (parallel processing) or sequential processing. DRS includes processing modules for layout analysis functions such as blob detection, block segmentation, and model matching, and for character recognition functions such as Kanji character recognition, Japanese postprocessing, postprocessing by a user, and error correction through a user interface. The character recognition functions on the card and the other processing-related recognition functions on the PC work cooperatively in the proposed framework. Within the basic framework, we have customized DRS for practical applications. Examples of successful applications—entry into a text database, creation of an electronic catalog, entry of family registration data, and entry of tag data in a manufacturing process—provide evidence of the processing accuracy and robustness of the framework.

Document image‐processing system for name and address recognition

This article describes a real-time document image processing system. Its objective is to recognize names and addresses from scanned address block images extracted from various tax forms of the United States Internal Revenue Service. The Name and Address Block Reader (NABR) system accepts both machine- and hand-printed address block images as input. Salient aspects of the system are presented, including document analysis (connected component analysis, address block extraction, label detection, hand-print/machine-print discrimination) and document recognition. Document recognition is performed in two nonidentical streams for machine-and hand-print; key steps are address parsing, character recognition, word recognition, and postal data base lookup (ZIP+4 and city-state-ZIP files). System output is a packet containing the results of recognition together with data base access status file. Real-time throughput (8500 forms/h) is achieved by employing a loosely coupled multiprocessing architecture where successive input images are distributed to available address recognition processors. The functional architecture, software design, system architecture, and the hardware implementation are described. Performance evaluation on machine- and hand-written addresses are presented. © 1996 John Wiley & Sons, Inc.

Document image-processing system for name and address recognition

This article describes a real-time document image processing system. Its objective is to recognize names and addresses from scanned address block images extracted from various tax forms of the United States Internal Revenue Service. The Name and Address Block Reader (NABR) system accepts both machine- and hand-printed address block images as input. Salient aspects of the system are presented, including document analysis (connected component analysis, address block extraction, label detection, hand-print/machine-print discrimination) and document recognition. Document recognition is performed in two nonidentical streams for machine-and hand-print; key steps are address parsing, character recognition, word recognition, and postal data base lookup (ZIP+4 and city-state-ZIP files). System output is a packet containing the results of recognition together with data base access status file. Real-time throughput (8500 forms/h) is achieved by employing a loosely coupled multiprocessing architecture where successive input images are distributed to available address recognition processors. The functional architecture, software design, system architecture, and the hardware implementation are described. Performance evaluation on machine- and hand-written addresses are presented. © 1996 John Wiley & Sons, Inc.

Font and Function Word Identification in Document Recognition

An algorithm is presented that identifies the predominant font in which the running text in an English language document is printed. Frequent function words (such asthe,of,and,a, andto) are also recognized as part of the font identification. Clusters of word images are generated from an input document and matched to a database of function words derived from fonts and document images. The font or document that matches best provides the identification of the predominant font and function words. This technique takes advantage of the fact that most machine-printed documents are prepared with a single predominant font. Also, the repeated words in the document are utilized to overcome noise in the input. Advantages of this technique include its use as a preprocessing step for a document recognition algorithm. Experimental results show high accuracy is achieved on a database of original and degraded document images.

Document recognition

Document recognition is a task in which a document in its physical presentation format is transformed into a structured author-oriented model of the document. The presentation format can be bitmaps of document pages, a description of the document in a Page Description Language (PDL), or encoding of the document in a printer or graphics language. The structured model is a format allowing for addition to the document, manipulation of the document, and reformating the layout and the output appearance of the document.Fully automatic document recognition is not possible, in general, for the same reason that it is not possible to de-translate computer programs automatically. However, it is possible to develop a man-assisted semi-automatic document recognition method. This method uses two passes. The first pass is completely automatic; it produces a document format called Interactive Document Model. The Interactive Document Model comprises recognized typesetting and descriptive structures together with derived ODA logical and layout structures for the document. The model generated in the first pass is enough for most purposes and applications. However, if it is not acceptable, the user can then enter the second pass and interactively edit the logical structure.This paper has three objectives. The first is to formalize the concept of document recognition. The second is to subdivide the problem of document recognition and classify it into a number of subproblems, each dealing with different aspects of the problem. The third objective is to introduce a problem which we wish to solve, and then to present a High Level Document Recognition method and the experience in developing and using a number of implementations of the method.

DRS: a workstation-based document recognition system for text entry

Document recognition system (DRS), a workstation-based prototype document analysis system that uses optical character recognition (OCR), is described. The system provides functions for image capture, block segmentation, page structure analysis, and character recognition with contextual postprocessing, as well as a user interface for error correction. All the functions except image capture and character recognition have been implemented by means of software for the Japanese edition of OS/2. >

First International Conference on Document Analysis and Recognition

First International Conference on Document Analysis and Recognition