A Description Method of Handprinted Chinese Characters

Imaging Science and Engineering Laboratory, Tokyo Institute of Technology, Midori-ku, Yokohama, Japan.
IEEE Transactions on Pattern Analysis and Machine Intelligence (Impact Factor: 4.8). 02/1979; DOI: 10.1109/TPAMI.1979.4766872
Source: IEEE Xplore

ABSTRACT A description method of handprinted Chinese characters is presented. In the method, a Chinese character is composed of some partial patterns which are constructed using the concatenate relation, cross relation, and near relation. The relations of relative location among partial patterns are used for categorization of the partial pattems. A Chinese character is expressed from the results of categorization.

  • [Show abstract] [Hide abstract]
    ABSTRACT: We describe a new method for representation of an image pattern, which we call the reduced special interval graph (RSIG). Any binary image pattern can be easily converted into a RSIG. Based on the RSIG, we can enlarge the scope of vision and process a much greater zone of the image. This helps us properly extract the feature of the image and identify the image pattern. We apply the RSIG to help us extract the strokes of Chinese characters by incorporating some knowledge about the structure of Chinese characters. We found this method can not only heuristically extract strokes but also heuristically eliminate noises including those added to strokes for the artistic sake. Experimental results reveal that the method based on the RSIG can extract the strokes of Chinese characters effectively and efficiently. Moreover, the RSIG can be applied to many other application areas, and it is an effective representation method in pattern recognition and image processing
    IEEE Transactions on Systems Man and Cybernetics 08/1995;
  • [Show abstract] [Hide abstract]
    ABSTRACT: A heuristic algorithm for the recognition of printed Chinese characters is presented. Preprocessing consists of identifying individual straight line primitive strokes of a Chinese character, and then identifying the sequence of occurrence of these primitive strokes in the course of two orthogonal and one diagonal scans. The results of the three scans are three ordered sets of primitive strokes that can be binary encoded. These three types of codes are called feature codes. The feature codes are used in the training phase and recognition phase by hashing. An experiment that trained on 13053 characters of a single font shows that only six pairs of characters have coincident feature codes. The recognition speed of this experiment is 44.4 milliseconds of 80386 CPU time per character (1,350 characters per minute excluding disk I/O time). The recognition rate is from 97.22% to 98.4%
    IEEE Transactions on Systems Man and Cybernetics 05/1995;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: A stroke feature distribution method is used to develop a scheme for the recognition of alphanumeric characters. The algorithm uses a gray-level image of the character as input. The image is normalized. The feature used is stroke length. The stroke feature vector is calculated at each point and averaged over image subdivisions which are equal in area. The results obtained for preclassification and for test sets of data are shown. All the algorithms were implemented in C on a MicroVax II workstation equipped with a high-resolution data translation image processing board.< >
    System Theory, 1988., Proceedings of the Twentieth Southeastern Symposium on; 04/1988