The IMAGE GROUP works to support the technology of image recognition in government and industry by developing new image recognition methods, developing techniques for the evaluation of existing methods, and providing technology transfer to the commercial imaging and document conversion industry. This group will also explore new work in face recognition, finger print classification, commercial areas of OCR, and visual aspects of document processing. The image recognition group is also organizing the First UNIPEN Benchmark of On-Line Handwriting Recognizers to determine the state of the art in pen-based handwriting recognition. All of these areas have been chosen to complement capabilities that exist at NIST, have application in large government or private markets, and address specific realizable applications.
Much of the data needed by computer users is in the form of visual information which is generated outside the computer but which would have greater value if it were labeled, indexed, and stored in the computer. For commercial application of this technology to succeed, the reliability and accuracy of computer image processing must be sufficient to allow large quantities of images to be processed with minimum human intervention. This technology will only succeed then in areas where it is fast, accurate, and invisible to potential users.
NIST develops evaluation methods for image based recognition systems. The image recognition group at NIST has pioneered many of these methods for developing test materials. As the size, speed, and accuracy of vision based recognition systems increases, automated methods for testing will continue to increase in importance.
NIST publishes data sets for testing systems that are the primary source for OCR applications and the sole source for fingerprints. NIST work has had a major impact on the OCR business. In the last four years, the image recognition group has produced 74 technical papers and reports, published 14 CDs and applied for eight patents, one of which was issued in of Feb. 1994.
The Image Group makes many files and documents available through its anonymous FTP server. The host sequoyah.nist.gov contains PostScript versions of the Group's Internal Reports, files associated with The First and Second Census Optical Character Recognition Systems Conferences, a catalog of character recognition and fingerprint matching databases available on CD-ROM, some source code, and more.
Created May 22, 2000.
Last modified September 21, 2000.
Contact
webmaster@magi.nist.gov with corrections/comments.