Word Image Matching in a Methodology for Degraded Text Recognition

A technique for the use of global context in text recognition is presented that determines equivalences between word images in a passage of text. Initial hypotheses for the identities of words are then generated by matching the word groups to language statistics that predict the frequency at which certain words will occur. This is followed by a recognition step and a relaxation-based control structure that resolves inconsistencies between several knowledge sources. This paper concentrates on the equivalence determination algorithm. A word matching technique is presented and its perfonnance .on a running text of about 1000 " word images is determined. Several levels of noise are introduced to simulate different J amounts of degradation introduced by fax machines or photocopiers. It is shown that the word matching algorithm maintains its ability to locate small groups of equivalent word images with high reliability even in the presence of noise.