Auto Recognition of Kanji Characters for Early-Modern Japanese Printed Books

日本国家饮食图书馆以互联网服务的形式向公众提供30万种近代印刷书籍。在这项研究中，我们开发了一个早期现代印刷汉字识别系统，以充分利用数字档案馆。该识别方法是脱机手写汉字识别方法的扩展。开发了一个门户网站，用于在用户的帮助下对系统进行额外学习。此外，我们还面临着几个问题，如出血和Rubi去除，以找到一些解决方案。

The national diet library of Japan provides the public with 300 thousands kinds of early-modern printed books as an internet service. In this research we develop an early-modern printed Kanji recognition system to make a full use of the digital archive. The recognition method is an extension of a recognition method for off-line handwritten Kanji characters. A portal site is developed for extra-learning the system with the help of users. Furthermore we faced several problems such as bleed through and rubi removal to find some solutions.