| 多种印刷体汉字识别的一种方法 |
| |
| 引用本文： | 朱学芳,毕厚杰,王绍棣.多种印刷体汉字识别的一种方法[J].南京邮电大学学报(自然科学版),1990(4). |
| |
| 作者姓名： | 朱学芳 毕厚杰 王绍棣 |
| |
| 作者单位： | 南京邮电学院无线电工程系，南京邮电学院无线电工程系，南京邮电学院无线电工程系 |
| |
| 摘 要： | 本文提出了一种新的描述多种印刷体汉字的方法——特征链码法.使用该方法无需先对汉字图像进行细化和平滑处理,可以克服汉字图像位移、不同大小以及一些噪声的影响.实验中针对常用的一级字库约3700个仿宋体、宋体、黑体(理论上还可以扩充到二级字库6700多个印刷体汉字,字体还可以增加楷体等),以及相应的尺寸出现的变化如24×24、24×48、48×48等做了特征抽取和识别试验,实验中还对链码特征库进行了几种预分类.
|
| 关 键 词： | 图像处理 汉字识别 特征抽取 |
A Method for Recognition of Multifont Printed Chinese Characters |
| |
| Zhu Xuefang Bi Houjie Wang Shaodi.A Method for Recognition of Multifont Printed Chinese Characters[J].Journal of Nanjing University of Posts and Telecommunications,1990(4). |
| |
| Authors: | Zhu Xuefang Bi Houjie Wang Shaodi |
| |
| Institution: | Department of Radio Engineering |
| |
| Abstract: | A new method for describing the features of multifont printed Chinese characters is proposed. It is called feature code method. It doesn't need thinning and smoothing the character image before abstracting its features. Small noise has little affect upon the recognition of multifont printed Chinese characters. Up-down shifting and left-right shifting images can also be recognised. The recognition rate: 100% (for learning sampler), recognition speed: 4.11s per character,on the average. Correct recognition rate: 99.53%, wrong recognition rate: 0.47% (for the multifont printed Chinese characters, in which there are small noises), the average speed: 4.22s per character (for Song Ti, Fang Song Ti,and Hei Ti about 3700 characters respectively). |
| |
| Keywords: | Image processing Chinese character recognition Feature extraction |
| 本文献已被 CNKI 等数据库收录！ |
