Method and apparatus for handwritten character recognition

• US 5,058,182 A
• Filed: 05/02/1988
• Issued: 10/15/1991
• Est. Priority Date: 05/02/1988
• Status: Expired due to Fees

• Current Assignee: The Research Foundation for The State University of New York (State University of New York)
• Original Assignee: The Research Foundation for The State University of New York (State University of New York)
• Inventors: Kuan, Chih-Chau, Hull, Jonathan J., Srihari, Sargur N.
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): Jung, Yon
• Application Number:
  US07/189,365
• Time in Patent Office: 1,261 Days
• Field of Search: 382/20, 382/22, 382/24, 382/28, 382/19, 382/25, 382/, 382/51, 382/13, 382/36
• US Class Current: 382/202
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
• CPC Class Codes: G06V 30/10 Character recognition

Abstract

A method and apparatus is provided for written character recognition. The method includes thresholding an input gray level character image to derive its bilevel image, extracting stroke-based features of said bilevel image, classifying the features based upon a set of predetermined rules, determining the identity of the character utilizing the classification, and indicating the identity of the character. A method and apparatus for extracting features of a written character is also provided. This method includes raster scanning the written character to determine stroke-based features, extracting vertical strokes of the character and extracting horizontal strokes of the character.

Claims

1. A method for extracting features of a written character, comprising:

raster scanning said written character to determine stroke-based features, including contour profiles;

extracting vertical strokes of said character; and

extracting horizontal strokes of said character, wherein said stroke-based features include a run length segment distribution spectrum which is used to estimate the character stroke width.

2. A method for extracting features of a written character, comprising:

raster scanning said written character to determine stroke-based features, including contour profiles;

extracting vertical strokes of said character; and

extracting horizontal strokes of said character, wherein said stroke-based features include a run length segment distribution spectrum which is used to estimate the character stroke width and also include a number of run length segments per horizontal scanning row.

3. A method for written character recognition, comprising:

thresholding an input gray level character to derive its bilevel image;

making a first-pass raster scan of said bilevel image and extracting stroke-based features of said image including a run length distribution spectrum which is used to estimate a stroke width of said character;

making a second-pass raster scan of said bilevel image and extracting vertical strokes of said image, where vertical strokes are defined as adjacent segments of the bilevel image that have stroke widths within the estimated stroke width and have approximately vertical collinear centers, and classifying any remaining components of the bilevel image as horizontal strokes which are segments which include true horizontal segments as well as segments having angular orientations;

classifying said extracted features and vertical and horizontal strokes based upon a set of predetermined rules;

determining the identity of said character utilizing said classification; and

indicating the identity of said character.

4. A method for written character recognition, comprising:

thresholding an input gray level character to derive its bilevel image;

making a first-pass raster scan of said bilevel image and extracting stroke-based features of said image, including separate right and left contour profiles, horizontal projection profile, run length distribution spectrum which is used to estimate a stroke width of said character, and number of run length segments per horizontal scanning row;

making a second-pass raster scan of said bilevel image and extracting vertical strokes of said image, where vertical strokes are defined as adjacent segments of the bilevel image that have stroke widths within the estimated stroke width and have approximately vertical collinear centers, and classifying any remaining components of the bilevel image as horizontal strokes which are segments which include true horizontal segments as well as segments having angular orientations;

classifying said extracted features and vertical and horizontal strokes based upon a set of predetermined rules;

determining the identity of said character utilizing said classification; and

indicating the identity of said character.