Document Image Analysis

Applications
  • Document Segmentation
  • Character Recognition
  • Vectorization
    • Projects
  • Document image analysis for map conversion
    • Publications
  • Publication list
  • Online Publications

    • Document Segmentation

    Original grey level image
    Result of global thresholding
    Result of adaptive thresholding

    A Fast Method for Adaptive Binarization

    In this study we have designed a new adaptive binarization method. The method works on subimages consisting of two floating windows; one larger window and one smaller placed at the center of the larger one. The windows are moved in steps equal to the size of the smaller window. A clustering of the pixels within the larger window is performed using Otsu's method for threshold selection. A test is used to decide whether the region consists of one or two classes, and the pixels within the smaller window are classified accordingly. Moving the windows in this fashin and using a clustering technique with properties like Otsu's method, we are able to retain the local contrast and obtain good binarizations of the images. The resultys of this method are compared to the best of a set of previously published and tested methods. On really difficult document images with large natural variations in contrast the performance of this method is significantly better. The method is also computationally fast.
    Eikvil, Line; Taxt, Torfinn and Moen, Knut:
    "A fast adaptive method for binarization of document images".
    1st International Conference on Document Analysis and Recognition (ICDAR). Proceedings. St. Malo, France September 30 - October 2, 1991.

    Character Recognition

    Decoding Bar Codes from Human-Readable Characters

    In this study we have decoded bar codes by recognizing the human-readable characters of the interpretation line printed below the bar pattern. Using this approach, we were able to successfully decode barcodes with a resolution of 0.8 pixels per module.
    Aas, Kjersti and Eikvil, Line:
    "Decoding bar codes from human-readable characters".
    Pattern Recognition Letters, Vol 18, 1997.

    Text recognition from grey level images using Hidden Markov models

    The problems of character recognition are today mainly due to imperfect thresholding and segmentation. In this paper a new approach to text recognition is presented which attempts to avoid these problems by working directly on grey level images and treating an entire word at the time. The features are found from the grey levels of the image, and a hidden Markov model is defined for each character. During recognition the most probable combination of models is found for each word by the use of dynamic programming.
    Aas, Kjersti; Eikvil, Line and Andersen, Tove:
    "Text recognition from grey level images using Hidden Markov models".
    Proceedings CAIP'95, Prague, Czech Republic September 1995.

    Automatic recognition of character strings in maps.

    In this study we describe tools for character string recognition on maps. Single character recognition is performed using elliptical Fourier descriptors applying a statistical classifier. The recognized characters are grouped into strings, and the syntax of these strings are then analysed to detect and correct errors. As training of the classifier is essential, tools for manual and automatic training and updating are included.
    Eikvil, Line; Aas, Kjersti and Holden, Marit:
    "Automatic recognition of character strings in maps".
    Proceedings CAIP'95, Prague, Czech Republic September 1995.

    Vectorization

    Tools for interactive map conversion and vectorization.

    The process of converting an analog map into structured digitized information requires several different operations, which are all time-consuming when performed manually. Strictly automatic processing is not always a possible solution, and an interactive approach can then be an alternative. This paper describes a tool for map conversion, focusing on the functionality for extraction of line structures. An interactive approach is used as it gives the user an opportunity to survey the process, and utilize human knowledge. The methods are based on contour following, extracting centre points needed for accurate vector representation of the line during tracing.
    Eikvil, Line; Aas, Kjersti; Koren, Hans:
    "Tools for Interactive Map Conversion and Vectorization".
    International Conference on Document Analysis and Recognition,
    Proceedings. Montreal No. Vol. 2, August 14-16, 1995.

    Back to the SAMBA department