| 摘要 |
PURPOSE: To execute appropriate pre-processing prior to character recognition processing. CONSTITUTION: The distortion of the gray scale picture of a document is corrected, and the gray scale picture is compaired with a threshold level to generate a binary picture from the distortion-corrected gray scale picture. Then the binary picture is subjected to segmentation processing to decide the position and the shape of individual character, and the gray scale picture information extracted for the individual character is subjected to recognition processing to decide the identification of the character, and to store the identification result of the character. Also, an under line is removed from a character with the under line in the picture to obtain the consecutive component in the binary picture, and plural sets of rules are applied to the consecutive component so as to filter the consecutive component of a text type from the consecutive component of a non-text type to subject only the consecutive component of the text type to character recognition processing. |
