Methods and systems for identifying text orientation in a digital image

• US 8,208,725 B2
• Filed: 06/21/2007
• Issued: 06/26/2012
• Est. Priority Date: 06/21/2007
• 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: Ferman, Ahmet Mufit, Speigle, Jon M.
• Primary Examiner(s): Carter, Aaron W
• Application Number:
  US11/766,640
• Publication Number:
  US 20080317343A1
• Time in Patent Office: 1,832 Days
• Field of Search: 382/168, 382177-188, 382/291, 715/204
• US Class Current: 382/177
• 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

Aspects of the present invention relate to systems and methods for determining text orientation in a digital image.

Claims

1. A method for determining a text orientation in a digital image, said method comprising:

in a first text line comprising a first plurality of text characters in a digital image, determining a first text-line orientation of said first text line, wherein said determining said first text-line orientation comprises;

determining a text-line bounding box for said first text line;

calculating an aspect ratio for said text-line bounding box; and

calculating said first text-line orientation based on said aspect ratio;

determining, for each of said text characters in said first plurality of text characters, a first-edge position measurement corresponding to a bounding edge associated with a first side of said first text line, thereby producing a plurality of first-edge position measurements;

determining, for each of said text characters in said first plurality of text characters, a second-edge position measurement corresponding to a bounding edge associated with a second side of said first text line, thereby producing a plurality of second-edge position measurements;

computing a first first-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said computing a first first-alignment feature comprises;

calculating a sample mean for said plurality of first-edge position measurements, thereby producing a ceiling measurement; and

calculating an error measure between said ceiling measurement and said plurality of first-edge position measurements, thereby producing said first first-alignment feature value;

computing a first second-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said computing a first second-alignment feature comprises;

calculating a sample mean for said plurality of second-edge position measurements, thereby producing a floor measurement; and

calculating an error measure between said floor measurement and said plurality of second-edge position measurements, thereby producing said first second-alignment feature value; and

determining a first text orientation of said first plurality of text characters in said digital image based on said first first-alignment feature value and said first second-alignment feature value, wherein said determining said first text orientation comprises determining a baseline-side of said first text line, wherein said determining said baseline-side of said first text line is based on the relative values of said first first-alignment feature value and said first second-alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language.

7. A method according to claim 1 further comprising:

in a second text line comprising a second plurality of text characters in said digital image, determining a second text-line orientation of said second text line;

computing a second first-alignment feature value relative to said second text-line orientation for said second text line in said digital image;

computing a second second-alignment feature value relative to said second text-line orientation for said second text line in said digital image;

determining a second text orientation of said second plurality of text characters in said digital image based on said second first-alignment feature value and said second second-alignment feature value; and

determining a page text orientation based on said first text orientation and said second text orientation.

9. A method according to claim 1 further comprising projecting said first plurality of text characters according to a skew angle prior to said determining, for each of said first plurality of text characters, said first-edge position measurement and prior to said determining, for each of said first plurality of text characters, said second-edge position measurement.

11. A method according to claim 1, wherein said determining a baseline-side of said first text line comprises:

associating said baseline-side of said first text line with said first side of said first text line when said first first-alignment feature value is less than said first second-alignment feature value; and

associating said baseline-side of said first text line with said second side of said first text line when said first second-alignment feature value is less than said first first-alignment feature value.

12. A method according to claim 1, wherein said determining a baseline-side of said first text line comprises:

receiving a language indicator for said digital image, wherein said language indicator indicates the language in which said first text line is written;

determining a relative likelihood of text characters with descenders to text characters with ascenders in said language;

associating said baseline-side of said first text line with said first side of said first text line when said first first-alignment feature value is less than said first second-alignment feature value and said relative likelihood indicates that text characters with descenders are less likely than text characters with ascenders;

associating said baseline-side of said first text line with said second side of said first text line when said first second-alignment feature value is less than said first first-alignment feature value and said relative likelihood indicates that text characters with descenders are less likely than text characters with ascenders;

associating said baseline-side of said first text line with said first side of said first text line when said second first-alignment feature value is greater than said first second-alignment feature value and said relative likelihood indicates that text characters with ascenders are less likely than text characters with descenders; and

associating said baseline-side of said first text line with said second side of said first text line when said second first-alignment feature value is less than said first and said relative likelihood indicates that text characters with ascenders are less likely than text characters with descenders.

13. A method according to claim 1, wherein said error measure between said floor measurement and said plurality of second-edge position measurements is an error measure selected from the group consisting of mean absolute error, mean-square error and root mean-square error.

14. A method according to claim 1, wherein said error measure between said ceiling measurement and said plurality of first-edge position measurements is an error measure selected from the group consisting of mean absolute error, mean-square error and root mean-square error.

16. A method according to claim 7, wherein said determining a second text orientation comprises:

determining a baseline-side of said second text line, wherein said determining said baseline-side of said second text line is based on the relative values of said second first-alignment feature value and said second second-alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language.

2. A method for determining a text orientation in a digital image, said method comprising:

in a first text line comprising a first plurality of text characters in a digital image, determining a first text-line orientation of said first text line;

determining, for each of said text characters in said first plurality of text characters, a first-edge position measurement corresponding to a bounding edge associated with a first side of said first text line, thereby producing a plurality of first-edge position measurements;

determining, for each of said text characters in said first plurality of text characters, a second-edge position measurement corresponding to a bounding edge associated with a second side of said first text line, thereby producing a plurality of second-edge position measurements;

orientation for said first text line in said digital image, wherein said computing first-alignment feature comprises;

calculating a sample mean for said plurality of first-edge position measurements, thereby producing a ceiling measurement; and

calculating an error measure between said ceiling measurement and said plurality of first-edge position measurements, thereby producing said first first-alignment feature value;

computing a first second-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said computing a first second-alignment feature comprises;

calculating a sample mean for said plurality of second-edge position measurements, thereby producing a floor measurement; and

calculating an error measure between said floor measurement and said plurality of second-edge position measurements, thereby producing said first second-alignment feature value;

determining a first text orientation of said first plurality of text characters in said digital image based on said first first-alignment feature value and said first second-alignment feature value; and

wherein said determining a first text-line orientation comprises;

determining a text-line bounding box for said first text line;

calculating an aspect ratio for said text-line bounding box; and

determining said first text-line orientation based on said aspect ratio.

6. A method according to claim 2 further comprising:

in a second text line comprising a second plurality of text characters in said digital image, determining a second text-line orientation of said second text line;

computing a second first-alignment feature value relative to said second text-line orientation for said second text line in said digital image;

computing a second second-alignment feature value relative to said second text-line orientation for said second text line in said digital image;

determining a second text orientation of said second plurality of text characters in said digital image based on said second first-alignment feature value and said second second-alignment feature value; and

determining a page text orientation based on said first text orientation and said second text orientation.

8. A method according to claim 2 further comprising projecting said first plurality of text characters according to a skew angle prior to said determining, for each of said first plurality of text characters, said first-edge position measurement and prior to said determining, for each of said first plurality of text characters, said second-edge position measurement.

10. A method according to claim 2, wherein said determining a first text orientation comprises:

determining a baseline-side of said first text line, wherein said determining said baseline-side of said first text line is based on the relative values of said first first-alignment feature value and said first second-alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language.

3. A method for determining a text orientation in a digital image, said method comprising:

in a first text line comprising a first plurality of text characters in a digital image, determining a first text-line orientation of said first text line, wherein said determining said first text-line orientation comprises;

determining a text-line bounding box for said first text line;

calculating an aspect ratio for said text-line bounding box; and

calculating said first text-line orientation based on said aspect ratio;

determining a first-side reference line for a first side of said first text line, said first-side reference line characterized by a first-side-reference-line position measurement;

determining a second-side reference line for a second side of said first text line, said second-side reference line characterized by a second-side-reference-line position measurement;

determining, for each of said first plurality of text characters, a first-edge position measurement corresponding to a bounding edge associated with a first side of said first text line, thereby producing a plurality of first-edge position measurements;

determining, for each of said first plurality of text characters, a second-edge position measurement corresponding to a bounding edge associated with a second side of said first text line, thereby producing a plurality of second-edge position measurements;

computing a first first-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said computing a first first-alignment feature comprises;

calculating a difference between each of said plurality of first-edge position measurements and said first-side-reference-line position measurement, thereby producing a first plurality of difference measurements;

calculating a first maximum, said first maximum corresponding to the maximum value of said first plurality of difference measurements;

calculating the absolute value of the difference between each of said first plurality of difference measurements and said first maximum, thereby producing a first plurality of difference-from-maximum values; and

summing said first plurality of difference-from-maximum values, thereby producing said first first-alignment feature value;

computing a first second-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said computing a first second-alignment feature comprises;

calculating a difference between each of said plurality of second-edge position measurements and said second-side-reference-line position measurement, thereby producing a second plurality of difference measurements;

calculating a second maximum, said second maximum corresponding to the maximum value of said second plurality of difference measurements;

calculating the absolute value of the difference between each of said second plurality of difference measurements and said second maximum, thereby producing a second plurality of difference-from-maximum values; and

summing said second plurality of difference-from-maximum values, thereby producing said first second-alignment feature value; and

determining a first text orientation of said first plurality of text characters in said digital image based on said first first-alignment feature value and said first second-alignment feature value, wherein said determining said first text orientation comprises determining a baseline-side of said first text line, wherein said determining said baseline-side of said first text line is based on the relative values of said first first-alignment feature value and said first second-alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language.

15. A method according to claim 3, wherein said first-side reference line corresponds to a first side of a bounding box for said first text line and said second-side reference line corresponds to a second side of said bounding box for said first text line.

17. A method according to claim 3, wherein said determining a baseline-side of said first text line comprises:

associating said baseline-side of said first text line with said first side of said first text line when said first first-alignment feature value is less than said first second-alignment feature value; and

associating said baseline-side of said first text line with said second side of said first text line when said first second-alignment feature value is less than said first first-alignment feature value.

18. A method according to claim 3, wherein said determining a baseline-side of said first text line comprises:

receiving a language indicator for said digital image, wherein said language indicator indicates the language in which said first text line is written;

determining a relative likelihood of text characters with descenders to text characters with ascenders in said language;

associating said baseline-side of said first text line with said first side of said first text line when said first first-alignment feature value is less than said first second-alignment feature value and said relative likelihood indicates that text characters with descenders are less likely than text characters with ascenders;

associating said baseline-side of said first text line with said second side of said first text line when said first second-alignment feature value is less than said first first-alignment feature value and said relative likelihood indicates that text characters with descenders are less likely than text characters with ascenders;

associating said baseline-side of said first text line with said first side of said first text line when said second first-alignment feature value is greater than said first second-alignment feature value and said relative likelihood indicates that text characters with ascenders are less likely than text characters with descenders; and

associating said baseline-side of said first text line with said second side of said first text line when said second first-alignment feature value is less than said first and said relative likelihood indicates that text characters with ascenders are less likely than text characters with descenders.

19. A method according to claim 3 further comprising projecting said first plurality of text characters according to a skew angle prior to said determining, for each of said first plurality of text characters, said first-edge position measurement and prior to said determining, for each of said first plurality of text characters, said second-edge position measurement.

4. A system for determining a text orientation in a digital image, said system comprising a non-transitory computer-readable medium comprising:

a text-line orientation determiner for determining a first text-line orientation of a first text line in a digital image, wherein said first text line comprises a first plurality of text characters;

a bounding-box determiner for determining a bounding box for each of said first plurality of text characters, thereby producing a plurality of bounding boxes, wherein each of said bounding boxes comprises;

a first edge, said first edge characterized by a first-edge position measurement, thereby producing a plurality of first-edge position measurements, and said first edge associated with a first side of said first text line; and

a second edge, said second edge characterized by a second-edge position measurement, thereby producing a plurality of second-edge position measurements, and said second edge associated with a second side of said first text line;

a first alignment feature calculator for computing a first-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein first-alignment feature calculator comprises;

a first sample-mean calculator for calculating a sample mean for said plurality of first-edge position measurements, thereby producing a ceiling measurement; and

a first error-measure calculator for calculating an error measure between said ceiling measurement and said plurality of first-edge position measurements, thereby producing said first first-alignment feature value;

a second alignment feature calculator for computing a second-alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said second-alignment feature calculator comprises;

a second sample-mean calculator for calculating a sample mean for said plurality of second-edge position measurements, thereby producing a floor measurement; and

a second error-measure calculator for calculating an error measure between said floor measurement and said plurality of second-edge position measurements, thereby producing said first second-alignment feature value;

a text orientation determiner for determining a text orientation of said first plurality of text characters in said digital image based on said first first-alignment feature value and said first second-alignment feature value, wherein said determining said text orientation comprises determining a baseline-side of said first text line, wherein said determining said baseline-side of said first text line is based on the relative values of said first-alignment feature value and said second-alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language; and

wherein said text-line orientation determiner comprises;

a text-line bounding box determiner for determining a text-line bounding box for said first text line;

an aspect-ratio calculator for calculating an aspect ratio for said text-line bounding box; and

wherein said text-line orientation determiner determines said text-line orientation based on said aspect ratio.

20. A system according to claim 4, wherein said text orientation determiner comprises:

a baseline-side determiner for determining said baseline-side of said first text line.

5. A system for determining a text orientation in a digital image, said system comprising a non-transitory computer-readable medium comprising:

a text-line orientation determiner for determining a first text-line orientation of a first text line in a digital image, wherein said first text line comprises a first plurality of text characters;

a text-line bounding box determiner for determining a first-text-line bounding box for said first text line, wherein said first-text-line bound box comprises;

a first-text-line first edge, said first-text-line first edge characterized by a first-text-line-first-edge position measurement and said first-text-line first edge associated with a first text-line-side of said first text line; and

a first-text-line second edge, said first-text-line second edge characterized by a first-text-line-second-edge position measurement and associated with a second text-line-side of said first text line;

a character-bounding-box determiner for determining a bounding box for each of said first plurality of text characters, thereby producing a plurality of bounding boxes, wherein each of said bounding boxes comprises;

a first edge, said first edge characterized by a first-edge position measurement, thereby producing a plurality of first-edge position measurements, and said first edge associated with a first side of said first text line; and

a second edge, said second edge characterized by a second-edge position measurement, thereby producing a plurality of second-edge position measurements, and said second edge associated with a second side of said first text line;

a first alignment feature calculator for computing a first alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said first alignment feature calculator comprises;

a first difference calculator for calculating a difference between each of said plurality of first-edge position measurements and said first-text-line first-edge position measurement, thereby producing a first plurality of difference measurements;

a first maximum calculator for calculating a first maximum, said first maximum corresponding to the maximum value of said first plurality of difference measurements;

a first absolute-value calculator for calculating the absolute value of the difference between each of said first plurality of difference measurements and said first maximum, thereby producing a first plurality of difference-from-maximum values; and

a first accumulator for summing said first plurality of difference-from-maximum values, thereby producing said first first-alignment feature value;

a second alignment feature calculator for computing a second alignment feature value relative to said first text-line orientation for said first text line in said digital image, wherein said second alignment feature calculator comprises;

a second difference calculator for calculating a difference between each of said plurality of second-edge position measurements and said first-text-line second-edge position measurement, thereby producing a second plurality of difference measurements;

a second maximum calculator for calculating a second maximum, said second maximum corresponding to the maximum value of said second plurality of difference measurements;

a second absolute-value calculator for calculating the absolute value of the difference between each of said second plurality of difference measurements and said second maximum, thereby producing a second plurality of difference-from-maximum values; and

a second accumulator for summing said second plurality of difference-from-maximum values, thereby producing said first second-alignment feature value; and

a text orientation determiner for determining a text orientation of said first plurality of text characters in said digital image based on said first first-alignment feature value and said first second-alignment feature value, wherein said determining said text orientation comprises determining a baseline-side of said first text line, wherein said determining said baseline-side of said first text line is based on the relative values of said first alignment feature value and said second alignment feature value and a relative frequency of occurrence of text characters with ascenders and text characters with descenders in a written language; and

wherein said text-line orientation determiner comprises;

a text-line bounding box determiner for determining a text-line bounding box for said first text line;

an aspect-ratio calculator for calculating an aspect ratio for said text-line bounding box; and

wherein said text-line orientation determiner determines said text-line orientation based on said aspect ratio.

21. A system according to claim 5, wherein said text orientation determiner comprises a baseline-side determiner for determining said baseline-side of said first text line.