Hierarchical recognition of structured hand-printed documents using rule-trees.

C. Cracknell, A. C. Downton, and L. Du
Department of Electronic Systems Engineering
University of Essex
Wivenhoe Park, Colchester  CO4 3SQ, U.K.
e-mail: craccb@@essex.ac.uk

Abstract:

Many computer vision applications exist where we wish to label objects which appear within an image. Possible relationships between different objects, and, the conditions under which a particular set of labels may or may not be applied are often known a-priori and encoded within a labelling model .

This paper overviews the use of a new modelling scheme (called a rule tree) which has been developed for facilitating hierarchical object labelling. Unlike other graph-based schemes which have been proposed, rule trees are very efficient: they permit labelling to be performed very quickly with negligible memory requirements. The use of rule trees is demonstrated by using the application domain of handwritten forms as an examplar.

• 1 Introduction.
• 2 Rule trees.
• 3 Implementation.
• 4 Results.
• 5 Conclusions.
• References