| 摘要 |
PURPOSE:To recognize characters in reference to a dictionary corresponding to a blurring process, by giving the weighted average to each feature element of an input character pattern with the contiguous elements and in response to the degree of complicacy of the input character pattern and then having the blurring process. CONSTITUTION:An optical input 1 given from a slip is fed to a photoelectric converting part 2, and a character estimated region is decomposed into (128X 128) picture elements. Then each picture element is converted into a binary digital signal. A coefficient deciding part 10 contains a table which shows the correspondence between the complicacy degree of an input character pattern and a blurring coefficient that is used to give a blurring process to the input character pattern. Thus the part 10 receives the complicacy degree of the input character pattern and decides the blurring coefficient based on the table of blurring coefficient. A blur processing part 11 receives a feature matrix of the input character pattern and the blurring coefficient and gives the weighted average to each feature element by means of two elements which are contiguous with each other in the direction orthogonal to the stroke direction. Thus the noticed feature element is blurred, and a blur feature matrix is produced. |
