【作者】 徐蔚然； 张洪刚； 郭军；

【Author】 Xu Weiran Zhang Honggang Guo Jun Beijing University of Posts & Telecom, Beijing, 100876

【机构】 北京邮电大学信息工程学院；

【摘要】 针对特定类别的图像,如何构造一种可以给出符合人类认识标准的处理结果的分割算法是很有意义的问题。本文提出一种把人类掌握的语法、语义信息对应关系转变成机器可以利用的用灰度分布特性表示的对应关系,从而构造出针对特定类别图像的二值化算法的可行思路。基于该思路,本文实际构造了几种针对不同银行票据图像的二值化算法。实验证明本文的方法简单而有效。更多还原

【Abstract】 How to construct special segmentation algorithms for particular kind of image which can give us a result that coincides with human’s understanding is a meaning work. A new pattern about constructing segmentation algorithms is put forward in this paper. First, we turn the corresponding relationship between syntactic information and semantic information that exists in humankind’s brain into the corresponding relationship that uses image’s features to denote. Then, we can construct a suitable segmentation algorithm. Based on this idea, several simple and effectual algorithms aimed at different kinds of financial document are practically constructed. Experiments show that our method is efficacious and simple.更多还原