一种离线手写体汉字切分的自适应算法

A Self-adaptive Segmentation Algorithm for Off-line Handwritten Chinese Characters

目前用于字符切分的算法主要有基于灰度直方图投影、字符连通域算法等,但这些算法不适用于相邻粘连汉字的切分。论文针对粘连汉字的切分,以邮件信函地址为对象,提出了一种离线手写体汉字切分的自适应算法。其基本步骤为:首先,基于灰值投影将手写汉字地址粗分为几个字段;其次,用傅立叶变换判断这些字段是否为粘连字段;再次,用汉字的字高和字宽的比值大小判断非粘连字是单字还是单字的部首;最后,以伸缩框法对粘连字分割,并对过分的部首进行合并。此算法的优势在于根据每个人书写的不同习惯,确定不同的伸缩框对汉字进行分割。

Several approaches such as image-projecting analysis and connected-components extraction algorithm are mostly applied in the field of the segmentation of handwritten chinese characters,which however is not quite available to divide the joined characters.To solve this problem,a self-adapive algorithm based on the projecting analysis,Fourier operators and flexible frame is put forward in this article.After the initial image processing,the coarse segments are af-firmed by means of projecting analysis.Subsequently the Fourier operator is used to give a verdict of whether the seg-ment is a joined couple characters or not.And at last a flexible frame is introduced to merge and divide the fine com-ponents which are processed and judged by the scale of the segments.The advantage of flexible frame can be benefit of the segmentation of different styles of handwriting charaxters,which is the emphasis of this article.