Character discrimination system employing height-to-width ratio and vertical extraction position information

• US 5,138,668 A
• Filed: 08/05/1991
• Issued: 08/11/1992
• Est. Priority Date: 05/19/1988
• Status: Expired due to Fees

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Abe, Keiko
• Primary Examiner(s): COUSO, JOSE L
• Application Number:
  US07/742,449
• Time in Patent Office: 372 Days
• Field of Search: 382/9-13, 382/34, 382/48
• US Class Current: 382/177
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/158 using character size, text spacings or pitch estimation

Abstract

A character recognition system which is improved in accuracy of integration for discrete characters. The system discriminate a component of any discrete character in accordance with a height-to-width ratio and a vertical extraction position of a rectangular area which is formed from a character row signal delivered to the system. Rectangular areas or character areas to be integrated are decided in accordance with an average character pitch of square or em characters.

Claims

1. A process of recognizing a character comprising the steps of:

a. electronically scanning a document upon which character information appears in rows and electronically generating a character row signal, which signal represents character string image data;

b. from the character row signal, electronically extracting non-discrete, ordinary characters and special characters in the form of em characters, and further electronically discriminating between discrete characters and components of discrete characters by;

i. defining a series of rectangular areas which circumscribe a complete character or a component of a discrete character;

ii. judging whether or not a selected one of the rectangular areas is a component of a discrete character by electronically comparing the ratio of the height of the selected rectangular area to the width of the selected rectangular area with a predetermined height-to-width ratio, and

iii. electronically comparing the vertical extraction position of the selected rectangular area to a predetermined vertical extraction position to thereby determine if the character within the selected rectangular area is a component of a discrete character;

c. from the results of steps a and b, generating electronic extracted character data representative of the extracted characters; and

d. electronically comparing the extracted character data with a dictionary of stored electronic standard character data, selecting as a recognized character a standard character whose data has the greatest similarity to an extracted character, and outputting electronic data corresponding to the recognized character.

2. A process of recognizing a character as recited in claim 1, wherein when it is electronically determined in steps ii and iii of step b of claim 1 that the first rectangular area circumscribes a component of a discrete character, step b further comprises the steps of:

iv. electronically determining from the character row signal whether or not a next adjacent rectangular area is another component of the discrete character by repeating steps ii and iii for the next adjacent rectangular area;

v. electronically determining whether or not the first rectangular area and the next adjacent rectangular area, if integrated together, would exceed a predetermined average character pitch and then electronically combining the next adjacent rectangular area with the first rectangular area to produce an integrated rectangular area if the character pitch of the integrated rectangular area does not exceed the average character pitch, and

vi. electronically discriminating whether or not the character information extracted from the integrated rectangular area are components of the discrete character by repeating steps ii through v while using the integrated rectangular area as the first rectangular area and subsequently outputting electronic, character image data as that of either a discrete character or as a component of a discrete character.

3. A process of recognizing a character as set forth in claim 1, wherein the predetermined height-to-width ratio is 1.

4. A process of recognizing a character as recited in claim 3, wherein the predetermined vertical extraction position is the vertical center of the character row.

5. An apparatus for recognizing a character comprising:

a. document reader means for electronically scanning a document upon which character information appears in rows and electronically generating a character row signal, which signal represents character string image data;

b. character extracting means supplied with the character row signal, for electronically extracting non-discrete, ordinary characters and special characters in the form of em characters, and further electronically discriminating between discrete characters and components of discrete characters by defining a series of rectangular areas which circumscribe a complete character or a component of a discrete character, judging whether or not a selected one of the rectangular areas is a component of a discrete character by comparing the ratio of the height of the selected rectangular area to the width of the selected rectangular area with a predetermined height-to-width ratio, and comparing the vertical extraction position of the selected rectangular area to a predetermined vertical extraction position to thereby determine if the character within the selected rectangular area is a component of a discrete character, and thereafter generating electronic extracted character data representative of the extracted characters; and

c. character discriminator means for electronically comparing the extracted character data with a dictionary of stored electronic standard character data, selecting as a recognized character a standard character whose data has the greatest similarity to an extracted character, and outputting electronic data corresponding to the recognized character.

6. An apparatus for recognizing a character as recited in claim 5, wherein when the character extracting means electronically determines that the first rectangular area circumscribes a component of a discrete character, the character extracting means further electronically determines from the character row signal whether or not a next adjacent rectangular area is another component of the discrete character by comparing the ratio of the height of the next adjacent rectangular area to the width of the next adjacent rectangular area with a predetermined height-to-width ratio and comparing the vertical extraction position of the next adjacent rectangular area to a predetermined vertical extraction position, electronically determines whether or not the first rectangular area and the next adjacent rectangular area, if integrated together, would exceed a predetermined average character pitch and then electronically combines the next adjacent rectangular area with the first rectangular area to product an integrated rectangular area if the character pitch of the integrated rectangular area does not exceed the average character pitch, and electronically discriminates whether or not the character information extracted from the integrated rectangular area are components of the discrete character by comparing the ratio of the height of the integrated rectangular area to the width of the integrated rectangular area with a predetermined height-to-width ratio and comparing the vertical extraction position of the integrated rectangular area to a predetermined vertical extraction position, and subsequently outputs electronic, character image data as that of either a discrete character or as a component of a discrete character.

7. An apparatus for recognizing a character as set forth in claim 5, wherein the predetermined height-to-width ratio is 1.

8. An apparatus for recognizing a character as recited in claim 7, wherein the predetermined vertical extraction position is the vertical center of the character row.