Data processing system and method for field extraction of scanned images of document forms

• US 5,416,849 A
• Filed: 10/21/1992
• Issued: 05/16/1995
• Est. Priority Date: 10/21/1992
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Huang, Shiow-Laang
• Primary Examiner(s): Mancuso, Joseph
• Application Number:
  US07/964,225
• Time in Patent Office: 937 Days
• Field of Search: 382/9, 382/61, 382/22, 382/44, 382/41, 395/100, 395/144, 395/145, 395/146
• US Class Current: 382/173
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1444 Selective acquisition, locating or processing of specific regions, e.g. highlighted text, fiducial marks or predetermined fields

Abstract

An improved data processing system and method are disclosed for field extraction of scanned images of document forms. The process makes use of a viewpoint structure which characterizes the interior and surrounding regions of a field from which an image is to be extracted for character recognition.

Claims

1. In a data processing system, a method for field extraction of scanned images of document forms, comprising the steps of:

creating a form definition for each form, which includes a form definition data set;

marking each field which is to be subjected to field extraction, on a master form;

defining a field viewpoint structure for each field, including field viewpoint primitive shapes including line ends, box ends, crossed lines, and blobs;

breaking the area within and surrounding a marked field into regions, an inner region enclosed by the marked field and eight outer regions which surround the marked field;

identifying types and locations of viewpoint primitives in each region;

linking the defined primitive into a view relation;

defining a field viewpoint structure as a set of all the viewpoint primitives and the view relation for a field;

storing the coordinates of the field mark and the field viewpoint structure for each field;

defining a collection of coordinates and field viewpoint structures for all fields on a form as a form definition data set for the form;

storing the form definition data set in the system until a particular filled in copy of the form is scanned into the system.

2. The method of claim 1 which further comprises the steps of:

scanning into the system a copy of a filled in form;

performing forms recognition to identify a form ID for the scanned-in form, which determines skew and offset of the scanned-in form;

using the form ID to access the form definition data set;

examining every field on the scanned-in form;

adjusting coordinates where each field is expected, by using the offset values for the scanned-in form;

comparing the stored viewpoint structure for a field in the data set, with the viewpoint structure of the field for the scanned-in form;

for each inner and surrounding region of a field, determining if there are any viewpoint primitives in the data set;

comparing viewpoint primitives in the data set with any viewpoint primitives in the corresponding region on the scanned-in form;

removing all matching viewpoint primitives from the scanned-in form image;

extracting field images from the fields after said removal step, and outputting the extracted field images to a character recognition process.