Return to Home Page

This book provides a background in the area of document image analysis.  It has general information on image analysis, information on document image analysis, and then information specific to the application of word and phrase recognition within document images.

Chapter 1 outlines some of the reasons documents are created as well as a bit of history behind document creation.  It discusses the different types of documents and the needs that each satisfies.

Chapter 2 discusses a collection of digital file formats, including vector based formats and image formats.  It discusses different compression techniques, character encoding, and digital image container formats as well as a few common file formats like the portable document format.  A description of the benefits and disadvantages of each format is presented, which is very useful for anyone in the area of document analysis and recognition since document images are built using these formats.

Chapter 3 provides details on message cryptography, verification, and legal aspects of document transfer.  This is useful for readers in the area of digital communications, with little relevance for document analysis and recognition performed offline.

Chapters 4 and 5 are the core of the discussion on document analysis and recognition.  Chapter 4 goes into detail about a selection of image processing algorithms used in document analysis and recognition.  This provides a good start for readers who are interested in document analysis and recognition but have no previous computer vision experience.  It also provides specific information on some preprocessing algorithms for document image analysis.  Chapter 5 starts by discussing methods of representing the document image, including the document object model (DOM).  The chapter then describes more preprocessing techniques useful for scanned documents.  Next, it provides a comparative review of the most common segmentation techniques in literature, stating their advantages and disadvantages.

Chapters 6 and 7 focus on processing text information extracted from the document image.  The chapters describe methods from the area of automated text understanding.  Chapter 6 describes methods for processing natural language, including parsing and sentence meaning.  Chapter 7 focuses on gathering specific information from the extracted text.  This includes searching for specific words as well as classifying the text. 

The book ends with two appendices, the first of which describes the Document Management Intelligent Universal System (DOMINUS).  This is a document image library system, which the author helped create.  The design of the system influenced the writing of the book.  The second appendix describes a few machine learning algorithms, including artificial neural networks, decision trees, k-nearest neighbor, inductive logic programming, naive Bayes, and Hidden Markov Models. 

For those working on document image analysis and recognition, chapters 4 and 5 are the most informative, along with Chapter 2. I would definitely recommend this book to novice researchers in document analysis and recognition, especially to those new to computer vision as well.