| 摘要 |
PROBLEM TO BE SOLVED: To provide a business form recognition system which is capable of quickly extracting a pair of item name and item value of a reading target not only out of a business form image in which a number of character strings are described but also out of a business form image containing dot-like noise. SOLUTION: After a character string region within a business form image is detected, a recognition priority of whether a character string before performing character string recognition is an item name or an item value of a reading target is set to that character string on the basis of image information of the character string (S110). Next, a character string recognition processing (S230), a dictionary collation processing (S240) and an item name-item value relationship extraction processing (S250) are performed successively from a character string rectangle having a high recognition priority. If it is determined that an item name-item value relationship of the reading target is extracted out of the business form image (S130), business form recognition processing is finished. Thus, business form recognition processing can be performed quickly without performing the character string recognition processing, the dictionary collation processing and the item name-item value relationship extraction processing on all the character strings within the business form image. COPYRIGHT: (C)2013,JPO&INPIT |
