Methods and Systems for Detecting Character Content in a Digital Image

• US 20080181496A1
• Filed: 01/26/2007
• Published: 07/31/2008
• Est. Priority Date: 01/26/2007
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Speigle, Jon M., Campbell, Richard John, Ferman, Ahmet Mufit
• Granted Patent: US 7,856,142 B2
• Time in Patent Office: Days
• US Class Current: 382/168
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18105 related to colour
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

Embodiments of the present invention comprise systems and methods for detecting glyph pixels in a rasterized image.

Claims

1. A method for classifying a pixel in a digital image, said method comprising:

a) determining a pixel of interest in a digital image;

b) evaluating a large-text-character condition for said pixel of interest and a first region substantially proximate to said pixel of interest, said first region comprising a first plurality of pixels and an associated first local background color estimate;

c) classifying said pixel of interest as a first-category pixel when said large-text-character condition is satisfied;

d) computing a dynamic threshold when said large-text-character condition is not satisfied, wherein said dynamic threshold is based on the color values in a second region substantially proximate to said pixel of interest, said second region comprising a second plurality of pixels and an associated second local background color estimate;

e) calculating a first similarity measure between said second local background color estimate and the color of said pixel of interest;

f) classifying said pixel of interest as a first-category pixel when said first similarity measure is greater than said dynamic threshold; and

g) classifying said pixel of interest as a second-category pixel when said first similarity measure is not greater than said dynamic threshold.

2. A method according to claim 1, wherein said computing a dynamic threshold comprises:

a) computing a distance between the color of each pixel in said second plurality of pixels and said second local background color estimate, thereby producing a plurality of color distances;

b) determining the range of said plurality of color distances;

c) determining the minimum-value color distance of said plurality of color distances;

d) adjusting said range of said plurality of color distances by a constant, thereby producing an adjusted range; and

e) adding said adjusted range to said minimum-value color distance, thereby producing said dynamic threshold.

3. A method according to claim 1, wherein said evaluating a large-text-character condition for said pixel of interest and said first region comprises:

a) computing a distance between the color of each pixel in said first plurality of pixels and said first local background color estimate, thereby producing a plurality of color distances;

b) determining an average of said plurality of color distances, thereby producing an first average color distance;

c) determining the range of said plurality of color distances;

d) comparing said average color distance to an average-distance threshold; and

e) comparing said range of said plurality of color distances to a color-range threshold.

4. A method according to claim 1, wherein said calculating a first similarity measure comprises calculating the distance between said second local background color estimate and the color of said pixel of interest.

5. A method according to claim 4, wherein said calculating the distance between said second local background color estimate and the color of said pixel of interest comprises a 3-dimensional city block distance calculation.

6. A method according to claim 1, wherein said first region and said second region are the same region.

7. A method according to claim 1, further comprising:

a) computing a distance between the color of each pixel in a third plurality of pixels comprising a third region substantially proximate to said pixel of interest and a third local background color estimate associated with said third region, thereby producing a plurality of color distances;

b) determining an average of said plurality of color distances;

c) evaluating said large-text-character condition only when said average of said plurality of color distances is not less than a contrast threshold; and

d) classifying said pixel of interest as a second-category pixel when said average of said plurality of color distances is less than said contrast threshold.

8. A method according to claim 1, further comprising:

a) computing a second similarity measure between a third local background color estimate for a third region substantially proximate to said pixel of interest;

b) evaluating said large-text-character condition only when said second similarity measure is larger than a similarity threshold; and

c) classifying said pixel of interest as a second-category pixel when said average of said second similarity measure is not larger than said similarity threshold.

9. A method according to claim 1, further comprising:

a) computing a distance between the color of each pixel in a third plurality of pixels comprising a third region substantially proximate to said pixel of interest and a third local background color estimate associated with said third region, thereby producing a plurality of color distances;

b) determining an average of said plurality of color distances;

c) evaluating said large-text-character condition only when said average of said plurality of color distances is not less than a contrast threshold or said third local background color estimate is equal to a page background color estimate; and

d) classifying said pixel of interest as a second-category pixel when said average of said plurality of color distances is less than said contrast threshold and said third local background color estimate is not equal to said page background color estimate.

10. A method according to claim 1, further comprising:

a) computing a second similarity measure between a third local background color estimate for a third region substantially proximate to said pixel of interest;

b) evaluating said large-text-character condition only when said second similarity measure is larger than a similarity threshold or said third local background color estimate is equal to a page background color estimate; and

c) classifying said pixel of interest as a second-category pixel when said average of said second similarity measure is not larger than said similarity threshold and said third local background color estimate is not equal to said page background color estimate.

11. A system for classifying a pixel in a digital image, said system comprising:

a) a pixel selector for determining a pixel of interest in a digital image;

b) a large-text evaluator for evaluating a large-text-character condition for said pixel of interest and a first region substantially proximate to said pixel of interest, said first region comprising a first plurality of pixels and an associated first local background color estimate;

c) a first classifier for classifying said pixel of interest as a first-category pixel when said large-text-character condition is satisfied;

d) a dynamic threshold calculator for computing a dynamic threshold when said large-text-character condition is not satisfied, wherein said dynamic threshold is based on the color values in a second region substantially proximate to said pixel of interest, said second region comprising a second plurality of pixels and an associated second local background color estimate;

e) a first similarity measure calculator for calculating a first similarity measure between said second local background color estimate and the color of said pixel of interest;

f) a second classifier for classifying said pixel of interest as a first-category pixel when said first similarity measure is greater than said dynamic threshold; and

g) a third classifier for classifying said pixel of interest as a second-category pixel when said first similarity measure is not greater than said dynamic threshold.

12. A system according to claim 11, wherein said dynamic threshold calculator comprises:

a) a distance calculator for computing a distance between the color of each pixel in said second plurality of pixels and said second local background color estimate, thereby producing a plurality of color distances;

b) a range calculator for determining the range of said plurality of color distances;

c) a minimum-value calculator for determining the minimum-value color distance of said plurality of color distances;

d) an adjustor for adjusting said range of said plurality of color distances by a constant, thereby producing an adjusted range; and

e) an adder for adding said adjusted range to said minimum-value color distance, thereby producing said dynamic threshold.

13. A system according to claim 11, wherein said large-text evaluator comprises:

a) a distance calculator for computing a distance between the color of each pixel in said first plurality of pixels and said first local background color estimate, thereby producing a plurality of color distances;

b) an average calculator for determining an average of said plurality of color distances, thereby producing an first average color distance;

c) a range calculator determining the range of said plurality of color distances;

d) a first comparator for comparing said average color distance to an average-distance threshold; and

e) a second comparator for comparing said range of said plurality of color distances to a color-range threshold.

14. A system according to claim 11, wherein said first similarity measure calculator comprises a distance calculator for calculating the distance between said second local background color estimate and the color of said pixel of interest.

15. A system according to claim 14, wherein said distance calculator comprises a 3-dimensional city-block distance calculation.

16. A system according to claim 11, wherein said first region and said second region are the same region.

17. A system according to claim 11, further comprising:

a) a color distance calculator for computing a distance between the color of each pixel in a third plurality of pixels comprising a third region substantially proximate to said pixel of interest and a third local background color estimate associated with said third region, thereby producing a plurality of color distances;

b) an average calculator for determining an average of said plurality of color distances;

c) said large-text evaluator evaluating said large-text-character condition only when said average of said plurality of color distances is not less than a contrast threshold; and

d) a fourth classifier for classifying said pixel of interest as a second-category pixel when said average of said plurality of color distances is less than said contrast threshold.

18. A system according to claim 11, further comprising:

a) a second similarity calculator for computing a second similarity measure between a third local background color estimate for a third region substantially proximate to said pixel of interest;

b) said large-text evaluator evaluating said large-text-character condition only when said second similarity measure is larger than a similarity threshold; and

c) a fourth classifier for classifying said pixel of interest as a second-category pixel when said average of said second similarity measure is not larger than said similarity threshold.

19. A system according to claim 11, further comprising:

a) a color distance calculator for computing a distance between the color of each pixel in a third plurality of pixels comprising a third region substantially proximate to said pixel of interest and a third local background color estimate associated with said third region, thereby producing a plurality of color distances;

b) an average calculator for determining an average of said plurality of color distances;

c) said large-text evaluator evaluating said large-text-character condition only when said average of said plurality of color distances is not less than a contrast threshold or said third local background color estimate is equal to a page background color estimate; and

d) a fourth classifier for classifying said pixel of interest as a second-category pixel when said average of said plurality of color distances is less than said contrast threshold and said third local background color estimate is not equal to said page background color estimate.

20. A system according to claim 11, further comprising:

a) a second similarity measure calculator for computing a second similarity measure between a third local background color estimate for a third region substantially proximate to said pixel of interest;

b) said large-text evaluator evaluating said large-text-character condition only when said second similarity measure is larger than a similarity threshold or said third local background color estimate is equal to a page background color estimate; and

c) a fourth classifier for classifying said pixel of interest as a second-category pixel when said average of said second similarity measure is not larger than said similarity threshold and said third local background color estimate is not equal to said page background color estimate.

21. A method for classifying a pixel in a digital image, said method comprising:

a) determining a pixel of interest in a digital image;

b) receiving a local background color estimate for a region substantially proximate to said pixel of interest, wherein said region comprises a plurality of pixels;

c) receiving a page background color estimate for said digital image;

d) calculating a similarity measure between said local background color estimate and the color of said pixel of interest, wherein said similarity measure comprising calculating the distance between the color of said local background color estimate and the color of said pixel of interest;

e) classifying said pixel of interest as a first-category pixel when said similarity measure is not greater than a similarity threshold and said local background color estimate is not equivalent to said page background color estimate;

f) calculating a contrast measure between a plurality of pixels in said region and said local background color when said similarity measure is greater than said similarity threshold or said local background color estimate is equivalent to said page background color estimate;

g) classifying said pixel of interest as a first-category pixel when said contrast measure is less than a contrast threshold and said local background color estimate is not equivalent to said page background color estimate;

h) evaluating a large-text-character condition for said pixel of interest when said contrast measure is not less than said contrast threshold or said local background color estimate is equivalent to said page background color estimate;

i) classifying said pixel of interest as a second-category pixel when said large-text-character condition is satisfied;

j) computing a dynamic threshold when said large-text-character condition is not satisfied, wherein computing said dynamic threshold comprises;

i) calculating a distance between the color of each of said plurality of pixels in said region and said local background color estimate, thereby producing a plurality of color distances;

ii) determining the range of said plurality of color distances;

iii) adjusting said range by a constant, thereby producing an adjusted range;

iv) determining the minimum-value color distance of said plurality of color distances; and

v) adding said adjusted range and said minimum-value color distance, thereby producing said dynamic threshold;

k) classifying said pixel of interest as a second-category pixel when said similarity measure is greater than said dynamic threshold; and

l) classifying said pixel of interest as a first-category pixel when said similarity measure is not greater than said dynamic threshold.