| 摘要 |
PURPOSE:To effectively sort the characters, by obtaining the complicatedness of the detailed part of a character pattern in terms of the structure. CONSTITUTION:A scanning photoelectric converting part 1 picks up and scans a character pattern to deliver the electric signal according to the concentration of an image. This electric signal is supplied to a preprocessing part 2 to be decomposed for its picture element by sampling and to be binary coded by quantization. Thus a sampled character pattern is formed to be stored and held in a pattern memory. A rough sorting dictionay 3 performs a feature recognizing process of a character pattern having features and reads selectively the recognition candidate character out of a candidate character memory 4 based on the result of the rough sorting. Based on the character pattern information stored and held in the pattern memory as well as the candidate character information, a recognizing part 5 recognizes said characters and delivers the result of recognition. The feature recognizing process is carried out by applying a weighted process related to the distance centering on a view point to the information of all sampling points when the view point is decided at an optional point in a matrix of MXN. |
