Digital Restoration of Archival Handwritten Texts¹

I. M. Bockstein**, N. A. Kuznetsov*, N. S. Merzlyakov*, L. I. Rubanov*, and G. A. Bogatova***

* Institute of Information Transmission Problems, Russian Academy of Sciences,
B. Karetnyi per. 19, Moscow, GSP-4, 101447 Russia;
tel.: +7(095)209-2883, fax: +7(095)209-0579; e-mail: nick@ippi.ac.msk.su

** Guzik Technical Enterprises, 4620 Fortran Dr., San Jose, CA 95134 USA;
tel.: (408)262-5500; fax: (408)263-2633; e-mail: bock@guzik.com

*** Institute of the Russian Language, Russian Academy of Sciences, ul. Volkhonka 18/2, Moscow, 121019 Russia

tel.: +7(095)202-0688; fax: +7(095)291-2317

e-mail: irl_ras@irl.msk.su

Abstract—The paper sums up a study designed to develop a practicable system for restoring and archiving
handwritten texts. The general structure of the system was set up. Methods are proposed whereby graphical
images can be automatically inputted into a computer and their quality can be enhanced. An efficient statistical
algorithm for adaptive image binarization was developed. Several new image-enhancing algorithms to handle
images of corrupted handwritten texts are presented. Options for compact storage of graphical images of hand
written texts are analyzed. It is shown that the best opportunity for their compact representation is offered by
the two-dimensional run-length coding technique. It was improved by individualized storage of codebooks for
“black” and “white” sequences of the image being coded. A method was developed for a compact representa
tion of codebooks. Facilities are examined for the organization of archival storage of handwritten texts. An algo
rithm is proposed to assure efficient interaction between the database management system and the image
archive. The algorithm retrieves data items from the archive, decodes them, and displays them on a screen.
Aprototype system was developed for digital restoration and archiving of handwritten texts and tailored to han
dle cards from a handwritten card file of the dictionary of Old Russian. The database includes information about
several hundred cards from the two-million card file. This database was recorded in the CD-ROM format and
can be replicated.

Pleiades Publishing home page | journal home page | top

If you have any problems with this server, contact webmaster.