Notes: The X-Y cut paper!

Available: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=602059

Abstract: A top-down page segmentation technique known as the recursive X-Y cut decomposes a document image recursively into a set of rectangular blocks. This paper proposes that the recursive X-Y cut be implemented using bounding boxes of connected components of black pixels instead of using image pixels. The advantage is that great improvement can be achieved in computation. In fact, once bounding boxes of connected components are obtained, the recursive X-Y cut is completed within an order of a second on Sparc-10 workstations for letter-sized document images scanned at 900 dpi resolution

Kieninger: Table Recognition based on Robust Block Segmentation

Embley, Hurst, Lopresti, Nagy (IJDAR): Table-processing paradigms: a research survey

Notes: an old paper, which deals with everything under the sun… what interests us is: (1) mentions the notion of a ‘zoomed-out view’; they use an OCR for scanned documents; postscript documents are used too (using a parser on top of ps2ascii) and a tree structure for the layout. The segmentation algorithm is based on, among other things, indentation alphabets and grammar-based (rule-based?) logical manipulations to generate a hierarchy (sounds too inflexible to me…) Table algorithm uses WDG and seems only to be limited to ASCII text (monospaced). Also includes some kind of DU for line drawings…

Conference: Mathematical and Computer Modelling, 1997

Available: http://www.dbai.tuwien.ac.at/education/wie/SS06/papers/94-Rus_Summers–Geometric_Algorithms_and_Experiments_for_Automated_Document_Structuring.ps.gz
(ps also available at citeseer)

Abstract: We present and analyse algorithms for the automated segmentation and classification of layout structures in electronic documents. The key idea is to use the patterns in the distribution of white space in a document to recognize and interpret its components. The segmentation algorithms classify these divisions as base-text, tables, indented lists, polygonal drawings and graphs. We present experimental data and discuss and information access application. Our methodology allows the automatic markup of documents (for instance in the SGML format) and the creation of multi-level indices and browsing tools for electronic libraries.

Notes: source is a scan; uses the traditional bitmap morphological etc. techniques; decision-tree (learning) based classification of blocks followed by knowledge-based detection of the layout structure (layout analysis) — declarative knowledge in prolog — followed by document classification and document understanding — a good paper to cite just for definitions!  finally, all this gets transformed into html/xml — but not as we know it (lixto)

Available: http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=38267FF42197FE109A013A6EE76EB09D?doi=10.1.1.33.4193&rep=rep1&type=pdf

Journal: IJDAR (2001)

Abstract: he transformation of scanned paper documents to a form suitable for an Internet browser is a complex process that requires solutions to several problems. The application of an OCR to some parts of the document image is only one of the problems. In fact, the generation of documents in HTML format is easier when the layout structure of a page has been extracted by means of a document analysis process. The adoption of an XML format is even better, since it can facilitate the retrieval of documents in the Web. Nevertheless, an effective transformation of paper documents into this format requires further processing steps, namely document image classification and understanding. WISDOM++ is a document processing system that operates in five steps: document analysis, document classification, document understanding, text recognition with an OCR, and transformation into HTML/XML format. The innovative aspects described in the paper are: the preprocessing algorithm, the adaptive page segmentation, the acquisition of block classification rules using techniques from machine learning, the layout analysis based on general layout principles, and a method that uses document layout information for conversion to HTML/XML formats. A benchmarking of the system components implementing these innovative aspects is reported.

Notes: Probably the first paper on DA of PDF.  Make use of the Adobe SDK.  Segmentation is via simple concepts based on the PDF objects (after pre-processing (line finding?) by the SDK).  Logical labelling: Mention blackboard architectures (would be an idea to look at a paper on these); for logical labelling uses knowledge database techniques.  For ‘document understanding’ (i.e. logical relationships — not sure if reading order too) uses rule-based heuristics. Now a long time ago!

Available: http://www.cs.nott.ac.uk/~dfb/Publications/Download/1995/will.pdf

Journal: Electronic Publishing, Vol. 8 (2 & 3), pp. 207-220 (June & Sept. 1995)

Summary: A strategy for document analysis is presented which uses Portable Document Format
(PDF—the underlying file structure for Adobe Acrobat software) as its starting point. This
strategy examines the appearance and geometric position of text and image blocks distributed
over an entire document. A blackboard system is used to tag the blocks as a first stage in
deducing the fundamental relationships existing between them. PDF is shown to be a useful
intermediate stage in the bottom-up analysis of document structure. Its information on line
spacing and font usage gives important clues in bridging the ‘semantic gap’ between the
scanned bitmap page and its fully analysed, block-structured form. Analysis of PDF can yield
not only accurate page decomposition but also sufficient document information for the later
stages of structural analysis and document understanding.

Notes: PDF!!! But they seem to fall back to a rendered version to perform segmentation, etc. This paper describes how the (single, full!) reading order of the front page of the International Herald Tribune is found. XMillum is mentioned

Available: http://diuf.unifr.ch/people/lalanned/Articles/XedDIAL04.pdf

Workshop: Document Image Analysis for Libraries, DIAL04

Abstract: PDF became a very common format for exchanging printable documents. Further, it can be easily generated from the major documents formats, which make a huge number of PDF documents available over the net. However its use is limited to displaying and printing, which considerably reduces the search and retrieval capabilities. For this reason, additional tools have recently appeared that allow to extract the textual content. However their practical use is limited in the sense that the text’s reading order is not necessary preserved, especially when handling multicolumn documents, or in presence of complex layout. Our thesis is that those tools do not consider the hidden layout and logical structures of documents, which could greatly improve their results. We propose a novel approach to overcome the document content extraction, by merging a) low-level extraction methods applied on PDF files with b) layout analysis performed on a synthetically generated TIFF image. The paper describes the various steps necessary to achieve this task. Finally, we present a first experiment on the restitution of the newspapers’ reading order which shows encouraging results.

deals with: everything you could possibly want to know about logical structure! (in more detail than i would need it, for example)  1998 vintage

Notes: unlike Chao’s previous paper, this paper doesn’t deal with document understanding per se, but rather examines the way PDFs are created from a variety of applications, e.g. Microsoft Word. A similar study should also be included in my thesis! Here, the Adobe SDK is used (which, I believe, performs some pre-processing — they talk of text runs). I need to do the same with PDFBox…

Available: http://www.science.uva.nl/events/dlia2001/program/s12_DL03.pdf

Workshop: Workshop on Document Layout Interpretation and its Applications (DLIA01)

Abstract: The Portable Document Format (PDF) has been
mostly used for posting the final form of documents. The
aim of our project is to analyze the layout, to modify the
layout or to re-use elements of PDF documents for
different media. Using PDFEdit in Adobe SDK, we built
tool to study the layout of documents and tool to select
page elements to compose a new page. We demonstrate
problems we encountered and propose possible solutions.

Notes: PDF!!! Uses a somewhat similar method for bottom-up ‘clustering’ to form segments (using e.g. line spacing and thresholds). Interesting heuristics which merge lines to form paragraphs and “flows”, without the requirement that they be rectangular.

Available: http://www.springerlink.com/index/B928PLAETK53AX91.pdf
direct download: http://www.springerlink.com/content/b928plaetk53ax91/fulltext.pdf (restricted); also on Google Books

Conference: DAS 2004

Abstract: Portable document format (PDF) is a common output format for electronic documents. Most PDF documents are untagged and do not have basic high-level document logical structural information, which makes the reuse or modification of the documents difficult. We developed techniques that identified logical components on a PDF document page. The outlines, style attributes and the contents of the logical components were extracted and expressed in an XML format. These techniques could facilitate the reuse and modification of the layout and the content of a PDF document page.

Available: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=156473 (restricted)

Notes: The majority of this paper deals with bitmap images and describes an OCR system with some basic logical structure detection. Although the paper claims only to deal with document analysis, some basic document understanding is performed, too. Topics of interest include: mentioning of top-down and bottom-up approaches (for segmentation) and the importance that the document be analysed on several levels with the possibility to accept or reject hypotheses at a later stage; binarization, morpohological operations, connected component methods, skew detection, line-finding (based on a pixel image), word/space detection and lots of stuff that doesn’t really concern us.

Abstract: The authors present a conceptual framework for solving the task of document analysis, which, in essence, consists in the conversion of the document’s pixel representation into an equivalent knowledge network representation holding the document’s content and layout. Starting on the pixel level, the formation of elementary geometric objects on which layout analysis as well as the definition of character objects is based is described. Character recognition accomplishes the mapping from geometric object to character meaning in ASCII representation. On the next level of abstraction words are formed and verified by contextual processing. Modeled knowledge about complete documents and about how their constituents are related to the application forms the highest level of abstraction. The various problems arising at each stage are discussed. The dependencies between the different levels are exemplified and technical solutions put forward

Journal: Proceedings of the IEEE
Publication Date: Jul 1992
Volume: 80, Issue: 7
On page(s): 1101-1119