基于AP聚类的中文字符分割

Chinese character segmentation based on AP clustering

中文字符的分割问题导致目前字符识别技术发展受限,对于图像模糊、中英文混排和字符粘连的文本图像,传统的投影分割等基于区域的分割方法分割效果差,导致文字识别率低。针对这些问题,本文将AP聚类算法应用于字符分割,该方法先将图像转化为灰度图,再将灰度图进行二值化处理,并对二值化图像进行除噪,然后根据聚类准则求解特征点之间的关系得到类中心点,最后根据类中心对特征点进行归类得到分割结果。实验结果表明,应用AP聚类方法能够取得比常规分割方法更好的分割效果。

At present, the development of character recognition technology is limited for character segmentation, the traditionalsegmentation method has poor segmentation effect on image blurring, mixed arranging of Chinese and English and characteradhesion. In order to overcome these questions, AP clustering algorithm in this paper is applied to character segmentation. Themethod is to transform the image into grayscale, be followed by greyscale image binarization processing, and get rid of the noise inbinary image. Then according to the clustering criterion, the relation between the feature points is obtained. Finally, the segmentationresults are obtained by using the classification of feature points according to the class center. The experiment results show that thismethod can achieve better segmentation effect than conventional segmentation method.