Document Mosaicing

Anthony Zappalá, Andrew Gee and Michael Taylor

University of Cambridge
Department of Engineering
Cambridge CB2 1PZ

Rank Xerox Research Centre
61 Regent Street
Cambridge CB2 1AB

{ 93arz,ahg } @eng.cam.ac.uk
mtaylor@cambridge.rxrc.xerox.com

Abstract:

Document scanning with an over-the-desk video camera offers an attractive alternative to the flat-bed scanner, photocopier or fax machine, since the documents need not be moved from their usual place of reading. However, standard CCD cameras can sample full A4 pages at only 50 dots per inch, which is not sufficient for anything more than the crudest reproduction. This motivates the use of ``mosaicing'', whereby multiple, overlapping images of the document are stitched together seamlessly to form a high resolution composite. Here we present a novel mosaicing system optimized for use with documents. The user simply slides the paper under the camera: all subsequent processing is fully automatic. The quality of the resulting images is sufficient for OCR or printing, though some post-processing is necessary.

• 1 Introduction
• 2 Obtaining suitable images
• 3 Detecting features for efficient matching
• 4 Establishing correspondences
• 5 Mosaicing the images
• 6 Conclusions and further work
• References