Methods and Systems for Identifying the Orientation of a Digital Image

• US 20090290801A1
• Filed: 05/23/2008
• Published: 11/26/2009
• Est. Priority Date: 05/23/2008
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Laboratories of America Incorporated (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Campbell, Richard John, Ferman, Ahmet Mufit
• Granted Patent: US 8,023,770 B2
• Time in Patent Office: Days
• US Class Current: 382/216
• CPC Class Codes: G06V 10/242 by image rotation, e.g. by 90 degrees

Abstract

Aspects of the present invention are related to systems and methods for determining the orientation of an electronic document image.

Claims

1. A method for determining the orientation of an image, said method comprising:

a) comparing a first connected-component object from an image to a first plurality of numeral templates, wherein said first plurality of numeral templates corresponds to a first orientation;

b) incrementing a first match counter, wherein said first match counter is associated with said first orientation, when said comparing a first connected-component object to a first plurality of numeral templates indicates a match between said first connected-component object and at least one numeral template of said first plurality of numeral templates;

c) comparing said first connected-component object to a second plurality of numeral templates, wherein said second plurality of numeral templates corresponds to a second orientation;

d) incrementing a second match counter, wherein said second match counter is associated with said second orientation, when said comparing said first connected-component object to a second plurality of numeral templates indicates a match between said first connected-component object and at least one numeral template of said second plurality of numeral templates; and

e) selecting an image orientation for said image based on said first match counter and said second match counter.

2. A method as described in claim 1 further comprising determining a reliability measure associated with said image orientation.

3. A method as described in claim 2 further comprising:

a) accepting said image orientation when said reliability measure meets a first condition; and

b) rejecting said image orientation when said reliability measure does not meet said first condition.

4. A method as described in claim 1, wherein said comparing a first connected-component object from an image to a first plurality of numeral templates comprises calculating a similarity measure between said first connected-component object and a first numeral template from said first plurality of numeral templates.

5. A method as described in claim 1 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component; and

b) normalizing said first binary connected component for size, thereby producing said first connected-component object.

6. A method as described in claim 1 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component;

b) normalizing said first binary connected component for size, thereby producing a first normalized, binary connected component;

c) processing said first normalized, binary connected component according to a distance transform, thereby producing said first connected-component object.

7. A method as described in claim 6, wherein said distance transform comprises a grassfire transform.

8. A method as described in claim 1, wherein said selecting an image orientation for said image based on said first match counter and said second match counter comprises:

a) selecting said first orientation when the value of said first match counter is greater than the value said second match counter; and

b) selecting said second orientation when the value of said first match counter is not greater than the value of said second match counter.

9. A method as described in claim 1 further comprising:

a) comparing said first connected-component object to a third plurality of numeral templates, wherein said third plurality of numeral templates corresponds to a third orientation;

b) incrementing a third match counter, wherein said third match counter is associated with said third orientation, when said comparing said first connected-component object to a third plurality of numeral templates indicates a match between said first connected-component object and at least one numeral template of said third plurality of numeral templates;

c) comparing said first connected-component object to a fourth plurality of numeral templates, wherein said third plurality of numeral templates corresponds to a fourth orientation;

d) incrementing a fourth match counter, wherein said fourth match counter is associated with said fourth orientation, when said comparing said first connected-component object to a fourth plurality of numeral templates indicates a match between said first connected-component object and at least one numeral template of said fourth plurality of numeral templates; and

e) selecting an image orientation for said image based on said first match counter, said second match counter, said third match counter and said fourth match counter.

10. A method as described in claim 9, wherein:

a) said second orientation is related to said first orientation by a 90°

clockwise rotation;

b) said third orientation is related to said first orientation by a 180°

clockwise rotation; and

c) said fourth orientation is related to said first orientation by a 270°

clockwise rotation.

11. A method as described in claim 9, wherein said selecting an image orientation for said image based on said first match counter, said second match counter, said third match counter and said fourth match counter comprises:

a) selecting said first orientation when said first match counter is greater than said second match counter, said third match counter and said fourth match counter;

b) selecting said second orientation when said second match counter is greater than said first match counter, said third match counter and said fourth match counter;

c) selecting said third orientation when said third match counter is greater than said first match counter, said second match counter and said fourth match counter; and

d) selecting said fourth orientation when said fourth match counter is greater than said first match counter, said second match counter and said third match counter.

12. A method as described in claim 1 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component;

b) determining if said first binary connected component is a numeral component; and

c) when said first binary connected component is determined to be a numeral component;

i) normalizing said first binary connected component for size, thereby producing a first normalized, binary connected component; and

d) processing said first normalized, binary connected component according to a distance transform, thereby producing said first connected-component object.

13. A method as described in claim 1 further comprising:

a) comparing a second connected-component object from an image to said first plurality of numeral templates;

b) incrementing said first match counter when said comparing a second connected-component object to said first plurality of numeral templates indicates a match between said second connected-component object and at least one numeral template of said first plurality of numeral templates;

c) comparing said second connected-component object to said second plurality of numeral templates; and

d) incrementing said second match counter when said comparing said second connected-component object to said second plurality of numeral templates indicates a match between said connected-component object and at least one numeral template of said second plurality of numeral templates.

14. A method for determining the orientation of an image, said method comprising:

a) determining a best match value, wherein said best match value comprises a match measure and an associated best-match orientation;

b) comparing a first connected-component object from an image to each numeral template in a first plurality of numeral templates, wherein said first plurality of numeral templates corresponds to a first orientation;

c) comparing said first connected-component object to each numeral template in a second plurality of numeral templates, wherein said second plurality of numeral templates corresponds to a second orientation;

d) updating said best match value based on said comparing a first connected-component object from an image to each numeral template in a first plurality of numeral templates and said comparing said first connected-component object to each numeral template in a second plurality of numeral templates;

e) incrementing a first match counter, wherein said first match counter is associated with said first orientation, when said best-match orientation is said first orientation;

f) incrementing a second match counter, wherein said second match counter is associated with said second orientation, when said best-match orientation is said second orientation; and

g) selecting an image orientation for said image based on said first match counter and said second match counter.

15. A method as described in claim 14 further comprising determining a reliability measure associated with said image orientation.

16. A method as described in claim 15 further comprising:

a) accepting said image orientation when said reliability measure meets a first condition; and

b) rejecting said image orientation when said reliability measure does not meet said first condition

17. A method as described in claim 14, wherein said comparing a first connected-component object from an image to each template in a first plurality of numeral templates comprises calculating a similarity measure between said first connected-component object and each template from said first plurality of numeral templates.

18. A method as described in claim 14 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component; and

b) normalizing said first binary connected component for size, thereby producing said first connected-component object.

19. A method as described in claim 14 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component;

b) normalizing said first binary connected component for size, thereby producing a first normalized, binary connected component;

c) processing said first normalized, binary connected component according to a distance transform, thereby producing said first connected-component object.

20. A method as described in claim 19, wherein said distance transform comprises a grassfire transform.

21. A method as described in claim 14, wherein said selecting an image orientation for said image based on said first match counter and said second match counter comprises:

a) selecting said first orientation when the value of said first match counter is greater than the value said second match counter; and

b) selecting said second orientation when the value of said first match counter is not greater than the value of said second match counter.

22. A method as described in claim 14 further comprising:

a) comparing said first connected-component object to each template in a third plurality of numeral templates, wherein said third plurality of numeral templates corresponds to a third orientation;

b) comparing said first connected-component object to each template in a fourth plurality of numeral templates, wherein said fourth plurality of numeral templates corresponds to a fourth orientation;

c) updating said best match value based on said comparing said first connected-component object from an image to each numeral template in a first plurality of numeral templates, said comparing said first connected-component object to each numeral template in a second plurality of numeral templates, said comparing said first connected-component object from an image to each numeral template in a third plurality of numeral templates and said comparing said first connected-component object from an image to each numeral template in a fourth plurality of numeral templates;

d) incrementing a third match counter, wherein said third match counter is associated with said third orientation, when said best-match orientation is said third orientation;

e) incrementing a fourth match counter, wherein said fourth match counter is associated with said fourth orientation, when said best-match orientation is said fourth orientation; and

f) selecting an image orientation for said image based on said first match counter, said second match counter, said third match counter and said fourth match counter.

23. A method as described in claim 22, wherein:

a) said second orientation is related to said first orientation by a 90°

clockwise rotation;

b) said third orientation is related to said first orientation by a 180°

clockwise rotation; and

c) said fourth orientation is related to said first orientation by a 270°

clockwise rotation.

24. A method as described in claim 22, wherein said selecting an image orientation for said image based on said first match counter, said second match counter, said third match counter and said fourth match counter comprises:

a) selecting said first orientation when said first match counter is greater than said second match counter, said third match counter and said fourth match counter;

b) selecting said second orientation when said second match counter is greater than said first match counter, said third match counter and said fourth match counter;

c) selecting said third orientation when said third match counter is greater than said first match counter, said second match counter and said fourth match counter; and

d) selecting said fourth orientation when said fourth match counter is greater than said first match counter, said second match counter and said third match counter.

25. A method as described in claim 14 further comprising:

a) detecting a first connected component in said image, thereby producing a first binary connected component;

b) determining if said first binary connected component is a numeral component; and

c) when said first binary connected component is determined to be a numeral component;

i) normalizing said first binary connected component for size, thereby producing a first normalized, binary connected component; and

ii) processing said first normalized, binary connected component according to a distance transform, thereby producing said first connected-component object.

26. A method as described in claim 14 further comprising:

a) comparing a second connected-component object from an image to each numeral template in said first plurality of numeral templates;

b) comparing said second connected-component object to each numeral template in said second plurality of numeral templates;

c) updating said best match value based on said comparing a second connected-component object from an image to each numeral template in said first plurality of numeral templates and said comparing said second connected-component object to each numeral template in said second plurality of numeral templates.

27. A method for generating numeral templates, said method comprising:

a) generating a plurality of binary, connected components representing a first numeral at a first orientation;

b) normalizing said plurality of binary, connected components to a first size, thereby producing a plurality of normalized components; and

c) averaging said plurality of normalized components.

28. A method for generating numeral templates, said method comprising:

a) generating a plurality of binary, connected components representing a first numeral at a first orientation;

b) normalizing said plurality of binary, connected components to a first size, thereby producing a plurality of normalized components;

c) distance transforming said plurality of normalized components, thereby producing a plurality of transformed, normalized components; and

d) averaging said plurality of transformed, normalized components.