| 摘要 |
The connected elements of an input image are obtained and grouped based on the relative positions of the connected elements and the similarity in thickness. Then, the character recognition level of a group is obtained by performing a character recognizing process. The obtained character recognition level is weighted by the area of a rectangular area. Using a total of the weighted values as an evaluation value of the group, the evaluation value is obtained for all combinations in all groups. The combination of the groups having the highest evaluation value is extracted as a character string.
|
