Character recognition method for recognizing character in an arbitrary rotation position

• US 5,031,225 A
• Filed: 12/09/1988
• Issued: 07/09/1991
• Est. Priority Date: 12/09/1987
• Status: Expired due to Fees

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Ishizaki, Hiroyoshi, Nakayama, Hiroshi, Tachikawa, Michiyoshi, Ishigami, Masayuki
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): Couso, Jose L.
• Application Number:
  US07/281,728
• Time in Patent Office: 942 Days
• Field of Search: 382/21, 382/22, 382/23, 382/9, 382/10, 382/46, 382/48
• US Class Current: 382/185
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1463 Orientation detection or correction, e.g. rotation of multiples of 90 degrees

Abstract

A character recognition method includes reading a document having characters displayed thereon and entering a document data representative of the content of the document. Image data of one of the characters from the document data is extracted. The one character has an orientation rotated by an arbitrary rotation angle from a predefined reference position for the one character. A first feature quantity of the one character is extracted from the extracted image data. The extracted feature quantity is converted into a second feature quantity of the one character in the reference position based on the arbitrary rotation angle. A dictionary which contains stored feature quantities of a plurality of characters in the reference position is interrogated so as to find a primary candidate character which has a feature quantity matching or most closely resembling the converted feature quantity.

Claims

1. A character recognition method comprising steps of:

reading a document having characters displayed thereon and entering document data representative of the content of the document;

extracting image data representative of one of the characters from the document data, said one character having an orientation rotated by an arbitrary rotation angle from a predefined reference position for said one character;

extracting a first feature quantity of said one character;

converting the extracted feature quantity into a second feature quantity of said one character in the reference position based on said arbitrary rotation angle;

referring to a dictionary which contains stored feature quantities of a plurality of characters in the reference position so as to find a primary candidate character which has a feature quantity matching or most closely resembling the second feature quantity; and

selecting said primary candidate character as said one character.

2. A character recognition method as claimed in claim 1 in which said step of converting the first feature quantity into the second feature quantity carries out a conversion dependent on a convention formula related to the relationship between the first feature quantity and the second feature quantity.

3. A character recognition method as claimed in claim 1 in which said step of extracting the first feature quantity includes steps of assigning direction codes to contour picture elements of said image data of said one character, each of said direction codes indicating a direction in which picture elements constituting said image data of said one character extend, and obtaining a histogram of the direction codes as said first feature quantity.

4. A character recognition method as claimed in claim 3 in which said step of converting the first feature quantity into the second feature quantity includes step of rearranging direction code components of the histogram based on said arbitrary rotation angle, said dictionary containing as the stored feature quantities histograms of the plurality of characters in the predefined reference position.

5. A character recognition method as claimed in claim 1 in which said step of extracting the first feature quantity includes steps of assigning direction codes to contour picture elements of the image data of said one character, each of said direction codes indicating a direction in which picture elements constituting the image data of said one character extend, dividing a character pattern of said one character into a plurality of regions, and obtaining a histogram of the direction codes as said first feature quantity with respect to each of the regions.

6. A character recognition method as claimed in claim 5 in which said step converting the first feature quantity into the converted feature quantity comprises step of rearranging direction code components of the histogram based on said arbitrary rotation angle with respect to each of the regions, said dictionary containing as the stored feature quantities histograms of the plurality of characters in the predefined reference position with respect to each of the regions.

7. A character recognition method as claimed in claim 1 in which said step of extracting the first feature quantity of said one character includes steps of assigning direction codes to contour picture elements of the image data of the input character in conformance with an m degree peripheral pattern, each of said direction codes indicating a direction in which picture elements constituting the image data of said one character extend, said m degree peripheral pattern describing a feature of a character in terms of distances to contour picture elements of the character along a scanning direction in each of a plurality of regions constituting a character pattern of said one character, and obtaining a histogram of the direction codes and the m degree peripheral pattern as said first feature quantity.

8. A character recognition method as claimed in claim 7 in which said step of converting the first feature quantity comprises step of rearranging direction code components of the histogram based on said arbitrary rotation angle, said dictionary containing as the stored feature quantities histograms and m degree peripheral patterns of the plurality of characters in the predefined reference position.

9. A character recognition method as claimed in claim 1 in which said arbitrary rotation angle is manually set in a register for use by said step of converting the first feature quantity into the second feature quantity.

10. A character recognition method as claimed in claim 1 in which said arbitrary rotation angle is selected from a group including 0, 90, 180 and 270 degrees counterclockwise from said predefined reference position.

11. A character recognition method as claimed in claim 1 in which said predefined reference position is a customary orientation of a character so that the character can be read in a normal manner.

12. A character recognition method as claimed in claim 1 which further comprises step of determining said arbitrary rotation angle and automatically setting said arbitrary rotation angle in a register for subsequent use by said step of converting the first feature quantity into the second feature quantity.

13. A character recognition method as claimed in claim 12 in which said step of determining said arbitrary rotation angle includes steps of carrying out a character recognition of said one character with respect to a plurality of rotation angles which are successively set in said register, and determining said array rotation angle by comparing the results of the character recognition carried out for each of the rotation angles set in said register.

14. A character recognition method as claimed in claim 12 in which said step of determining said arbitrary rotation angle includes steps of carrying out a character recognition of said one character with respect to a plurality of rotation angles which are successively set in said register, detecting differences between features of said one character and the primary candidate character from results of the character recognition carried out with respect to each of the rotation angles set in said register, and determining said arbitrary rotation angle by detecting the primary candidate character exhibiting a smallest difference.

15. A character recognition method as claimed in claim 12 in which said step of determining said arbitrary rotation angle includes steps of carrying out a character recognition of said one character with respect to a plurality of rotation angles which are successively set in said register, and determining said arbitrary rotation angle by comparing the results of the character recognition carried out for each of the rotation angles set in said register by referring to at least a key character or a key word likely to be found in the document.

16. A character recognition method as claimed in claim 12 in which said step of determining said arbitrary rotation angle includes steps of carrying out a character recognition of the input character with respect to a plurality of rotation angles which are successively set in said register, and determining said arbitrary rotation angle by comparing the results of the character recognition carried out for each of the rotation angles set in said register by referring to at least a character or word stored in a dictionary.

17. A character recognition method as claimed in claim 12 in which said step of converting the first feature quantity includes step of carrying out a conversion dependent on a conversion formula relates to the relationship between the first feature quantity and the second feature quantity.

18. A character recognition method comprising steps of:

reading an original document having characters displayed thereon;

storing document data representative of the character content of the document;

extracting image data representative of one of said characters from said document data, said one character having a first orientation;

extracting a firs feature quantity of said one character from said image data;

converting said first feature quantity into a second feature quantity based on a predetermined relationship between said first orientation and a second predefined orientation for said one character that is different from said first orientation;

referring to a dictionary which stores feature quantities of a plurality of characters in said second orientation so as to fund a primary candidate character that has a feature quantity matching or closely resembling said second feature quantity; and

selecting said primary candidate character as said one character.

19. A character recognition method as claimed in claim 18, wherein said predetermined relationship is a rotation angle indicative of the relative difference between said one character in said first and second orientations.

20. A character recognition method as claimed in claim 18, in which said step of converting the first feature quantity into the second feature quantity carries out a conversion dependent on a conversion formula related to the relationship between the first feature quantity and the second feature quantity.

21. A character recognition method as claimed in claim 18, in which said step of extracting said first feature quantity includes steps of assigning direction codes to contour picture elements of said image data of said one character, each of said direction codes indicating a direction in which picture elements constituting said image data of said one character extend, and obtaining a histogram of the direction codes as said first feature quantity.

22. A character recognition method as claimed in claim 21, in which the step of converting the first feature quantity into the second feature quantity further comprises rearranging direction code components of the histogram based on said predetermined relationship, said dictionary containing as the stored feature quantities predefined histograms of the plurality of characters in the second orientation.

23. A character recognition method as claimed in claim 18, in which said step of extracting the first feature quantity includes steps of assigning direction codes to contour picture elements of the image data of said one character, each of said direction codes indicating a direction in which picture elements constituting the image data of said one character extend, dividing a character pattern of said one character into a plurality of regions, and obtaining a histogram of the direction codes as said first feature quantity with respect to each of the regions.

24. A character recognition method as claimed in claim 23, in which said step of converting the first feature quantity further comprises rearranging direction code components of the histogram based on said predetermined relationship with respect to each of the regions, said dictionary containing as the stored feature quantities histograms of the plurality of characters in the second orientation with respect to each of the regions.

25. A character recognition method as claimed in claim 19, in which said rotation angle is manually set in a register for used by said converting step for converting the first feature quantity into the second feature quantity.

26. A character recognition method as claimed in claim 19, in which said arbitrary rotation angle is selected from a group consisting of 0, 90, 180 and 270 degrees counterclockwise from said predefined reference position.

27. A character recognition method as claimed in claim 18, which further comprises step of determining said rotation angle and automatically setting said rotation angle in a register for subsequent use in said converting step for converting the first feature quantity into the second feature quantity.

28. A character recognition method as claimed in claim 27, in which said step of determining said rotation angle includes steps of carrying out a character recognition of said one character with respect to a plurality of rotation angles which are successively set in said register, detecting differences between features of said one character and the primary candidate character from results of the character recognition carried out with respect to each of the rotation angles set in said register, and determining said rotation angle by detecting the primary candidate character exhibiting the smallest difference.

29. A character recognition method as claimed in claim 27, in which said step of determining said rotation angle includes steps of carrying out a character recognition of said one character with respect to a plurality of rotation angles which are successively set in said register, and determining said arbitrary rotation angle by comparing the results of the character recognition carried out for each of the set rotation angles by referring to at least a key character or a key word likely to be found in the document.

30. A character recognition method as claimed in claim 27, in which said step of determining said rotation angle includes steps of carrying out a character recognition of the input character with respect to a plurality of rotation angles which are successively set in said register, and determining said rotation angle by comparing the results of the character recognition carried out for each of the set rotation angles by referring to at least a character or word stored in a dictionary.