Abstract
Now we are using efficiently many database systems of technical papers, geographical maps, software and so on. As databases especially concerned with publishing and printing field, book database systems of companies, universities or individuals are used. Tedious and laborious jobs, however, are required to make such database systems. In order to assist in making them automatically, we are developing a pattern recognition method to extract and recognize characters on the shelfback of a book, i.e., title, author's name and publishing company, from a bookshelf image. As the first step of this research, the present paper describes an image processing method to divide a bookshelf image into several books, and extract character regions on the shelfback of each book. This method is also applied to other application, such as automatic checking of book position in a library.
