| 摘要 |
PROBLEM TO BE SOLVED: To improve matching efficiency and to provide high-accuracy character recognition by finding out a character image corresponding to the high-reliability character image of the recognized result preserved to the preceding line. SOLUTION: In the processing of the first line, character segmenting processing is executed, the character in the same shape is extracted and preserved in the same image memory 105 and character recognition is performed later (S203-S205). Continuously, several kinds of post-processing such as language processing is executed while referring to a language dictionary 104, the recognized result is preserved in a recognized result buffer 106, and the high-reliability character image is preserved in the same image memory 105 together with the matching result (S206 and S207). In the processing of lines after the second line, character segmenting processing is performed similarly to the first line, plural segment candidates are extracted and before matching processing, the character image in the same shape as the character image preserved in the same image memory 105 is found out of the segment candidates. Thus, partial processing is omitted and processing is accelerated.
|
