Research on Automatic Character Recognition of Wooden Blocked Tibetan Manuscripts

藏文木版藏文刻本中有许多重要的佛经文本，应用面向对象技术有望有效地对其进行研究。这项研究的目的是为了方便将这些木刻手稿整理成罗马文字，以便在当今先进的计算机辅助下促进佛教文献的研究。我们尝试用欧几里得距离和尊重权重来识别木刻藏文手稿中的相似字符。本面向对象方法对530个闭合数据的识别率达到了96%。为了实现藏文字符的自动识别，从木版藏文手稿中分离出每个音节是非常重要的。然而，75%的木刻汉字都有相连的结构。从而达到了70%的字符切分率。对于印刷体藏文字符，目前的面向对象方法已经达到了99.5%的识别率。首先，使用图像扫描仪将图像数据数字化，并将其带入个人计算机，并自动识别。自动识别一个藏文字符大约需要5秒。我们目前正在研究将这种已经开发的方法应用于藏文手稿的识别。

There are a number of important Buddhist texts written in Tibetan wooden blocked manuscripts which are expected to be efficiently studied by applying Object Oriented Technology. This research has originated from the desire to facilitate the work in cording and compiling these wooden blocked manuscripts into Romanized form to encourage Buddhist literature studies by present day advanced computer assistance.Many Tibetan characters are similar with each other. We try that similar characters are recognized themselves for wooden blocked Tibetan manuscripts by using Euclidean distance with deferential weight. A 96 % recognition rate has been achieved by the present Object Oriented procedure for 530 closed data. It is important to separate each syllable from wooden block Tibetan manuscripts to recognize automatically the Tibetan characters. However, 75 % of wooden blocked characters have connected structures. Therefore, 70 % character segmentation rate is achieved. For printed Tibetan characters, a 99.5 % recognition rate has already been achieved by present Object Oriented procedure. First, image data is digitized and taken into a personal computer by using an image scanner, and automatically recognized. It is taken about 5 seconds to recognize one Tibetan character automatically. We are presently studding to apply this already developed method to recognize the block Tibetan manuscripts.