| 摘要 |
PURPOSE:To perform a correct character segmentation and to enhance the recognition rate by using character segmentation information the front and the back lines when segmenting the character every one character from a vertical projection. CONSTITUTION:A character segmentation part 3 segments the character every one character based on vertical projection information formed in a vertical projection forming part on a character image of a line buffer. In a character segmentation position memory part 4, a left end position and a right end position of the respective characters when segmenting the character in a front line are stored. In the segmentation part 3, by using the position information of the memory part 4, the character segmentation is changed, and final character segmentation position is decided to segment the character every one character. A recognition part 5 extracts the characteristic of the segmented character pattern and collates to a characteristic dictionary to output the recognition result. Thereby, the correct character segmentation can be performed and the recognition rate can be enhanced.
|
