Method and apparatus for vectorizing documents and symbol recognition

• US 4,759,075 A
• Filed: 04/10/1986
• Issued: 07/19/1988
• Est. Priority Date: 03/14/1983
• Status: Expired due to Fees

• Current Assignee: Intergraph Corporation (Hexagon AB)
• Original Assignee: Ana-Tech Corporation
• Inventors: Lipkie, Curtis A., Kleca, Eugene A.
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/850,232
• Time in Patent Office: 831 Days
• Field of Search: 382/13, 382/41, 382/44, 382/46, 382/34
• US Class Current: 382/288
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: H04N 1/4115 involving the recognition of specific patterns, e.g. by symbol matching
• CPC Class Codes: H04N 1/417 using predictive or differential encoding

Abstract

The invention in a preferred embodiment recognizes symbols by determining the center of mass and maximum extremity of a symbol candidate, and comparing it to a reference library after normalizing with respect to scale, orientation and center of mass.

Claims

1. A method of creating a normalized representation of a symbol candidate appearing in a two-dimensional image field comprising the steps of:

scanning the candidate in an ordered manner and assigning to the candidate a representation by coordinates along axes of a coordinate system, the coordinates corresponding to the line and position values relating to the scan signals generated by the candidate;

determining the center of mass of the representation of the candidate;

determining a point of the candidate at a maximal distance from its center of mass;

scaling the candidate with respect to its center of mass by a factor proportional to the reciprocal of the distance from the center of mass to the point of maximum extremity;

rotating the candidate around its center of mass so that the segment determined by its center of mass and the point of maximum extremity lies at a prescribed angular orientation with respect to the axes of the coordinate system;

comparing the scaled rotated candidate to a library of stored representations of symbols and determining the one with which it has the greatest degree of overlap.

2. A system for identifying, from a cluster of vector data obtained from scanning graphic information along a direction of scan, a symbol in a library of candidates for the symbol, the system comprising:

means for determing the center of mass of the cluster;

means for determining a point of the cluster at a maximal distance from its center of mass;

means for scaling the cluster with respect to the center of mass by a factor proportional to the reciprocal of the distance from the center of mass to the point of maximum extremity, so as to normalize the cluster;

means for computing a representation of the cluster equivalent to rotating the cluster around the center of mass so that the segment determined by the center of mass and the point of maximum extremity lies at a prescribed angular orientation with respect to the direction of scan; and

means for comparing the scaled rotated candidate to a library of stored representations of symbols and determining the one with which it has the greatest degree of overlap.