Selective Rescanning for OCR Systems
Original Publication Date: 1980-Mar-01
Included in the Prior Art Database: 2005-Feb-13
In optical character recognition (OCR), there are approximately 10% to 15% of failure patterns (rejects/substitutions) which are very close to being acceptable by the OCR system. Usually, there are a few bits in the pattern which are too many, or not sufficient, or slightly out of order which describes the pattern, that causes the pattern to be an OCR failure. The distribution of such irregularities within the pattern is random and requires substantial engineering changes in order to turn some of these failures into accepts. An alternative solution to these marginal patterns is to rescan them while they are still present in the scanner. The information extraction (digitalization) by the scanner is random to a degree, and pattern rescanning can recover the marginal distortions within the pattern.