UESHIMA Shinichi Kansai University, Faculty of Letters, Associate Professor, 文学部, 助教授 (10184920)
SHIBAI Keiji Kansai University, Faculty of Letters, Associate Professor, 文学部, 助教授 (00144311)
NAGATA Hidemasa Shiga University, Faculty of Education, Professor, 教育学部, 教授 (90027532)
|Budget Amount *help
¥3,900,000 (Direct Cost: ¥3,900,000)
Fiscal Year 1989: ¥900,000 (Direct Cost: ¥900,000)
Fiscal Year 1988: ¥3,000,000 (Direct Cost: ¥3,000,000)
This research consists of the following three parts:(I) survey of the advance of image database systems;(2) the design and construction of the image database system of the Dun-Huang Han wooden strips, and indexing of ancient vocabularies found in the wooden strips;(3) image processing of the images of wooden strips and feature extraction by introducing geometrical parameters.
Firstly, we survey image database systems constructed in the various fields, especially focusing on the organization of systems, and point out the underlying problems from the viewpoint of their possibility of the future advance in the human sciences.
Secondly, we design and construct an experimental image database system of the Dun-Huang Han wooden strips. All the vocabularies implemented in the system are based on both our interpretation of sentences in the wooden strips and many materials presented in this field, while their image data are obtained from the documents of E. Chavannes and H. Maspero through a scann
er or a CCD camera. We explain the way how to cut out a single image automatically from many wooden strips placed in the same page, and how to cut out a single character semi-automatically in a row of Chinese characters, in detail. We have indexed several kinds of vocabularies, such as person's names, place names, government official names, names of era, idioms. By using this system we are able to retrieve a single Chinese character from each sentence in a wooden strip, in the style of KWIC(Key Word In Context) sorting, as well as in the usual format. We believe that this system gives a foundation in the research of Han wooden strips.
Thirdly, we convert the original Chinese character images into two-gray level images and thinner the resulting images by well-known image processing techniques so that we may extract features of ancient Chinese characters and get a key to rebind wooden strips in the original order. We examine geometrical parameters, by applying them to "Lin-hu-sui ce-shu', "ce-shu of Yuan-Kang 5', "Ju-Yan Han-Jian ce-shu", which were rebound for their original order in 1987, 1961, 1988, respectively. Less