System and method for locating a character set in a digital image

• US 7,630,544 B1
• Filed: 04/06/2005
• Issued: 12/08/2009
• Est. Priority Date: 04/06/2005
• Status: Expired due to Fees

• Current Assignee: Seiko Epson Corporation (Seiko Group)
• Original Assignee: Seiko Epson Corporation (Seiko Group)
• Inventors: Zhou, Hui
• Primary Examiner(s): Ge; Yuzhen
• Application Number:
  US11/099,951
• Time in Patent Office: 1,707 Days
• Field of Search: None
• US Class Current: 382/164
• CPC Class Codes: G06V 20/62 Text, e.g. of license plates, overlay texts or captions on TV images
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18105 related to colour

Abstract

A method of locating a character set in a digital image comprises examining pixels of the digital image to detect pixels having specified color attributes. Regions of the detected pixels are identified and the identified regions are clustered according to color similarity. Each cluster is examined to detect the cluster meeting specified morphological constraints thereby to locate the character set. An apparatus for locating a character set in a digital image is also provided.

Claims

1. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein during pixel examining, the pixels are examined to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties; and

wherein detected pixels having said specific color attributes are those having red, green and blue channel values outside of the following constraints;

red<

80;

green>

180 and red<

100;

blue>

180 and red<

100;

red>

100 and green>

100 and blue<

80;

red>

180 and green>

180 and blue>

180;

green>

(red+50) and red>

100;

blue>

(red+50) and red>

100;

green<

(red+15) and red<

100;

blue>

(red+15) and red<

100;

|red-green|<

30 and |red-blue|<

30 and |green-blue|<

30;

where each of red, green, and blue is from 0 to 255.

2. The method of claim 1, wherein detected pixels having said specific color attributes are those pixels having red, green and blue channel values outside of the following constraints:

red<

80;

green>

180 and red<

100;

blue>

180 and red<

100;

red>

100 and green>

100 and blue<

80;

red>

180 and green>

180 and blue>

180;

green>

(red+50) and red>

100;

blue>

(red+50) and red>

100;

green<

(red+15) and red<

100;

blue>

(red+15) and red<

100;

|red-gree|<

30 and |red-blue|<

30 and |green-blue|<

30;

where each of red, green, and blue is from 0 to 255; and

having hue, saturation and luminance channel values outside of the following constraints;

80<

hue<

280;

saturation<

0.28;

luminance<

0.40;

where hue is from 0 to 360, saturation is from 0 to 1.0, and luminance is from 0 to 1.00.

3. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein during pixel examining, the pixels are examined to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties; and

wherein detected pixels having said specific color attributes are those pixels having hue, saturation and luminance channel values outside of the following constraints;

80<

hue<

280;

saturation<

0.28;

luminance<

0.40;

where hue is from 0 to 360, saturation is from 0 to 1.0, and luminance is from 0 to 1.00.

4. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein during pixel examining, the pixels are examined to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties;

wherein the pixels are examined to determine if the pixels meet both the specified RGB and HSL properties; and

wherein the pixels of said digital image are processed in parallel to determine pixels meeting both the specified RBG and HSL properties.

5. The method of claim 4 further comprising filtering and blurring said digital image prior to said examining thereby to remove noise in said digital image.

6. The method of claim 5, wherein detected pixels having said specific color attributes are those pixels having red, green and blue channel values outside of the following constraints:

red<

80;

green>

180 and red<

100;

blue>

180 and red<

100;

red>

100 and green>

100 and blue<

80;

red>

180 and green>

180 and blue>

180;

green>

(red+50) and red>

100;

blue>

(red+50) and red>

100;

green<

(red+15) and red<

100;

blue>

(red+15) and red<

100;

|red-green|<

30 and |red-blue|<

30 and |green-blue|<

30;

where each of red, green, and blue is from 0 to 255; and

having hue, saturation and luminance channel values outside of the following constraints;

80<

hue<

280;

saturation<

0.28;

luminance<

0.40;

where hue is from 0 to 360, saturation is from 0 to 1.0, and luminance is from 0 to 1.00.

7. The method of claim 5 wherein said specified RGB properties assist in detecting yellowish or reddish pixels.

8. The method of claim 4 wherein at least one of the RGB and HSL properties and the morphological constraints are tunable.

9. The method of claim 4 wherein each of the RGB and HSL properties and the morphological constraints is tunable.

10. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set; and

wherein during said identifying, parts of the digital image with little blue information are focused on to identify said pixel regions; and

wherein during said identifying, red and blue channel images are extracted from said digital image and processed to yield a multi-level grey scale image, successive pixels of common grey scale forming said regions.

11. The method of claim 10 wherein regions including pixels at edges of said grey scale image are removed.

12. The method of claim 11 wherein said multi-level grey scale image includes four grey scale levels.

13. The method of claim 12 wherein during said processing a difference image of said red and blue channel images is formed, said difference image being posterized to yield said four level grey scale images.

14. The method of claim 10 further comprising examining the sizes of said regions to detect regions within a specified size range thereby to detect potential characters.

15. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein said pixel regions are clustered according to hue;

wherein during said clustering the average hue of the pixels in each region is detected and wherein said regions are clustered according to average hue;

wherein said regions are clustered into one of three hue levels to yield first, second and third hue channel images, said hue channel images being examined to detect the cluster of regions meeting said specified morphological constraints; and

wherein during said clustering the hue of the pixels is examined to determine if a threshold number of pixels has a hue level falling within a specified range, if so, all of the pixels being used to form said first hue channel image, said second and third hue channel images being set to black.

16. The method of claim 15 wherein said threshold number is 85% of the pixels.

17. The method of claim 15 wherein if the threshold number of pixels is not reached, the pixels having a hue falling within said specified range being used to form said first hue channel image, the remaining pixels being examined to determine if a second threshold number of pixels has a hue level falling within a second specified range, if so, all of the remaining pixels being used to form said second hue channel image and said third hue channel image being set to black.

18. The method of claim 17 wherein said second threshold number is 85%.

19. The method of claim 17 wherein if the second threshold number of pixels is not reached, the pixels having a hue falling within said second specified range being used to form said second hue channel image, the remaining pixels being used to form said third hue channel image.

20. The method of claim 19, further comprising:

performing erosion and dilation operations on said hue channel images.

21. The method of claim 20 wherein said morphological constraints require the cluster to have at least three horizontally aligned characters of similar height.

22. The method of claim 21 wherein said morphological constraints further require all of the characters in said cluster to be bound by a rectangle between 20 and 80 pixels.

23. The method of claim 22 wherein said morphological constraints further require the rectangle to have a width between 20 and 90% of the digital image width.

24. The method of claim 23, wherein when more than one cluster meets said morphological constraints, the bounding rectangle encompassing the cluster having the greatest number of regions is deemed to be the cluster representing said character set.

25. A method of locating a character set in a digital image comprising:

using a processing unit to perform the following steps;

examining pixels of said digital image to detect pixels having specified color attributes;

identifying regions of said detected pixels;

clustering said identified regions according to color similarity; and

examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein said morphological constraints require the cluster to have at least three horizontally aligned characters of similar height;

wherein said morphological constraints further require all of the characters in said cluster to be bound by a rectangle between 20 and 80 pixels; and

wherein said morphological constraints further require the rectangle to have a width between 20% and 90% of the digital image width.

26. The method of claim 25 wherein when more than one cluster meets said morphological constraints, the bounding rectangle encompassing the cluster having the greatest number of regions is deemed to be the cluster representing said character set.

27. An apparatus for locating a character set in a digital image comprising:

a color attribute detector for examining pixels of said digital image to detect pixels having specified color attributes;

a pixel color similarity detector identifying regions of pixels having said specified color attributes and clustering identified regions according to color similarity; and

a morphological constraint applier examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein said color attribute detector examines the pixels to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties;

wherein the color attribute detector examines the pixels to determine if the pixels meet both the specified RGB and HSL properties; and

wherein detected pixels having said specific color attributes are those pixels having red, green and blue channel values outside of the following constraints;

red<

80;

green>

180 and red<

100;

blue>

180 and red<

100;

red>

100 and green>

100 and blue<

80;

red>

180 and green>

180 and blue>

180;

green>

(red+50) and red>

100;

blue>

(red+50) and red>

100;

green<

(red+15) and red<

100;

blue>

(red+15) and red<

100;

|red-green|<

30 and |red-blue|<

30 and |green-blue|<

30;

where each of red, green, and blue is from 0 to 255; and

having hue, saturation and luminance channel values outside of the following constraints;

80<

hue<

280;

saturation<

0.28;

luminance<

0.40;

where hue is from 0 to 360, saturation is from 0 to 1.0, and luminance is from 0 to 1.00.

28. An apparatus for locating a character set in a digital image comprising:

a color attribute detector for examining pixels of said digital image to detect pixels having specified color attributes;

a pixel color similarity detector identifying regions of pixels having said specified color attributes and clustering identified regions according to color similarity; and

a morphological constraint applier examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein said color attribute detector examines the pixels to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties;

wherein the color attribute detector examines the pixels to determine if the pixels meet both the specified RGB and HSL properties; and

wherein said color attribute detector processes the digital image in parallel to determine pixels meeting the specified RGB and HSL properties.

29. An apparatus according the claim 28 further comprising a filter to filter and blur the digital image prior to examining by said color attribute detector.

30. An apparatus for locating a character set in a digital image comprising:

a color attribute detector for examining pixels of said digital image to detect pixels having specified color attributes;

a pixel color similarity detector identifying regions of pixels having said specified color attributes and clustering identified regions according to color similarity; and

a morphological constraint applier examining each cluster to detect the cluster meeting specified morphological constraints thereby to locate said character set;

wherein said color attribute detector examines the pixels to determine if the pixels meet at least one of specified red, green, blue (RGB) color properties and specified hue, saturation and luminance (HSL) properties;

wherein the color attribute detector examines the pixels to determine if the pixels meet both the specified RGB and HSL properties;

wherein said pixel color similarity detector focuses on parts of the digital image with little blue information to identify the pixel regions; and

wherein the pixel color similarity detector extracts red and blue channel images from the digital image and processes the red and blue channel images to yield a multi-level grey scale image with successive pixels of common grey scale forming said regions.

31. An apparatus according to claim 30 wherein said multi-level grey scale image includes four grey scale levels and wherein said pixel color similarity detector forms a difference image of said red and blue channel images and posterizes the difference image to yield the four level grey scale image.

32. An apparatus according to claim 31 wherein said pixel color similarity detector detects the average hue of the pixels in each region and clusters the regions according to average hue.

33. An apparatus according to claim 32 wherein the color pixel similarity detector clusters the regions into one of three hue levels to yield first, second and third hue channel images, the hue channel images being examined by the morphological constraint applier to locate said character set.

34. An apparatus according to claim 33 wherein said morphological constraint applier examines clusters to detect clusters having at least three horizontally aligned characters of similar height.

35. An apparatus according to claim 34 wherein said morphological constraint applier further examines the clusters to detect clusters bounded by a rectangle between 20 and 80 pixels.

36. An apparatus according to claim 35 wherein said morphological constraint applier further examines the clusters to detect a rectangle having a width between 20% and 90% of the digital image width.

37. An apparatus according to claim 36 wherein said morphological constraint applier deems the bounding rectangle encompassing the cluster having the greatest number of regions as that representing the character set when more than one cluster meets the morphological constraints.