Fast Key-In for Machine-Printed OCR-based Systems

• US 20090245659A1
• Filed: 03/31/2008
• Published: 10/01/2009
• Est. Priority Date: 03/31/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Tzadok, Asaf, Walach, Eugeniusz
• Granted Patent: US 8,103,132 B2
• Time in Patent Office: Days
• US Class Current: 382/224
• CPC Class Codes: G06V 10/987 with the intervention of an operator

Abstract

A method for correcting results of OCR or other scanned symbols that initially scans a document, performs OCR classification on the scanned document. Clustering of characters/symbols classifications that results from the OCR based on shapes is followed by the creation of super-symbols based on at least first differences in the shapes of the clustered characters/symbols exceeding a first threshold. A carpet of the super symbols is displayed for analysis testing and that provides a display that emphasizes localized differences in similar symbols. Depending on the results of the analysis testing, the carpet of super symbols is one of: (1) stored as a cluster symbols when the carpet of super-symbols passes all of the analysis testing; (2) used to create additional super symbols based on at least a second difference in the shapes of clustered symbols exceeding a second threshold and returning the additional super symbols to the displaying step for further analysis testing; and (3) rejected as a clustered of symbols when the super symbols fail most of the analysis testing and stored by manually keying-in the symbols.

Claims

1. A method for correcting results of optical character recognition (OCR) comprising:

scanning a document image;

performing OCR classification on the document image;

clustering symbols from the OCR classification based on shapes of the symbols;

creating super symbols based on first differences in the shapes of the clustered symbols exceeding a first threshold and that provides a display that emphasizes localized differences in similar symbols;

displaying a carpet of super symbols for analysis testing;

storing the clustered symbols when the carpet of super symbols passes all of the analysis testing;

creating additional sub groups of super symbols based on second differences in the shapes of clustered symbols exceeding a second threshold and returning to displaying step when the carpet of super symbols passes most of the analysis testing;

rejecting clustered symbols when the carpet of super symbols fails most of the analysis testing; and

storing rejected clustered symbols by manually keying-in the symbols.