| 摘要 |
PURPOSE:To easily and quickly classify an inputted character by extracting a skeleton of the character as features from a character pattern inputted as a black-and-white binary picture. CONSTITUTION:The black-and-white binary character pattern from a character pattern memory 3 is inputted to a longitudinal run length measuring means 5A and a transverse run length measuring means 5B to measure the length of the longest run in each scanning, and a longitudinal threshold and a transverse threshold are determined based on this length information by a longitudinal threshold determining means 6A and a transferse threshold determining means 6B. Areas where runs longer than the longitudinal threshold and the transverse threshold exist are extracted as a longitudinal skeleton and a transverse skeleton by a longitudinal skeleton extracting means 7A and a transverse skeleton extracting means 7B and a longitudinal quantized skeleton and a transverse quantized skeleton are extracted by a longitudinal skeleton quantizing means 8A and a transverse skeleton quantizing means 8B. These skeletons are stored in a skeleton memory 4 as features of the character pattern.
|
