Method and apparatus for character recognition

• US 5,680,479 A
• Filed: 04/24/1992
• Issued: 10/21/1997
• Est. Priority Date: 04/24/1992
• Status: Expired due to Term

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Wang, Shin-Ywan, Vaezi, Mehrzad R., Sherrick, Christopher Allen
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Bella, Matthew C.
• Application Number:
  US07/873,012
• Time in Patent Office: 2,006 Days
• Field of Search: 382/13, 382/9, 382/22, 382/56, 382/54, 382/36, 382/48, 382/37, 382/18, 382/170, 382/171, 382/176, 382/177, 382/178, 382/191, 382/194, 395/148
• US Class Current: 382/176
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/15 Cutting or merging image elements, e.g. region growing, watershed or clustering-based techniques
• CPC Class Codes: G06V 30/155 Removing patterns interfering with the pattern to be recognised, such as ruled lines or underlines
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text

Abstract

In a character recognition system or the like, method and apparatus for selecting blocks of pixels from pixel image data so as to permit identification and grouping of similarly-typed pixels, such as text-type pixels and non-text-type pixels. Pixel image data is inputted and, if the pixel image data is not binary image data then the pixel image data is converted into binary pixel image data. Blocks of pixel image data are selected by outlining contours of connected components in the pixel image data, determining whether the outlined connected components include text unit or non-text units based on the size of the outlined connected components, selectively connecting text units widthwisely to form text lines based on proximity of adjacent text units, and selectively connecting text lines vertically to form text blocks based on proximity of adjacent text lines and on the position of non-text units between text lines. A hierarchical tree is formed based on the outlined connected components.

Claims

1. A computer-implemented method of selecting blocks of pixels from pixel image data comprising, in order, the steps of:

outlining contours of connected components in the pixel data;

forming a rectangle around each connected component outlined in said outlining step;

forming a hierarchical tree based on the outlined connected components, and designating as a descendent a connected component which is within a rectangle formed around another connected component;

a first connecting step in which rectangles are selectably connected widthwisely based on size and proximity to other rectangles to form text lines;

a second connecting step in which a plurality of text lines formed in the first connecting step are selectably connected vertically based on size and proximity to other formed text lines to form text blocks, at least one formed text block having a plurality of formed text lines; and

modifying the hierarchical tree based on the first and second connecting steps.

2. A method according to claim 1, further comprising the step of inputting pixel image data, wherein said input step includes the step of converting the pixel image data into binary pixel image data in the event that the pixel image data is not binary pixel image data.

3. A method according to claim 1, wherein in said forming step the smallest rectangle is formed around each connected component.

4. A method according to claim 1, further comprising the step of forming a hierarchical tree of the connected components outlined in said outlining step based on the position of corresponding rectangles formed in said forming step.

5. A method according to claim 4, further comprising the step of classifying the rectangles formed in said forming step into text and non-text units.

6. A method according to claim 5, further comprising the step of recognizing character images in blocks of text units.

7. A method according to claim 5, wherein said classifying step is performed in accordance with a predetermined threshold of unit height.

8. A method according to claim 5, wherein said classifying step is performed in accordance with statistical information concerning the height of said rectangles formed in said forming step.

9. A method according to claim 5, wherein said first connecting step and said second connecting step are not performed for non-text units.

10. A method according to claim 5, further comprising the step of deriving white contours for non-text units.

11. A method according to claim 10, wherein non-text units are designated as tale units in accordance with the number of white contours.

12. A method according to claim 10, further comprising the step of calculating the white contour filling rate.

13. A method according to claim 12, wherein the non-text unit is not designated as image data in the event that the filling rate is high.

14. A method according to claim 12, further comprising the step of recombining non-grid-arranged white contours.

15. A method according to claim 14, wherein the non-text unit is designated as a table unit in the event that a rate of recombination in said recombining step is not high.

16. A method according to claim 10, wherein white contours are calculated in 4-directions.

17. A method according to claim 1, wherein contours of connected components are outlined in at least 8-directions.

18. A method according to claim 1, wherein in said outlining step, contours of connected components are outlined only at exterior portions of the connected components.

19. A method according to claim 1, further comprising the step of detecting gaps in the pixel image data, and wherein in said first connecting step rectangles are not connected into lines in the event that a gap separates the rectangles.

20. A method according to claim 19, wherein columns are detected in accordance with vertically extending gaps between rectangles.

21. A method according to claim 1, wherein said second connecting step includes the step of determining non-text boundaries between lines of text data connected in said first connecting step, and wherein said second connecting step does not connect lines vertically into blocks in the event that there is an intervening non-text boundary.

22. A method according to claim 1, further comprising the step of compressing the pixel image data before said outlining step.

23. A method according to claim 1, further comprising the step of classifying as text and non-text:

(i) hierarchical tree components; and

(ii) their descendents.

59. A method according to claim 1, further comprising the step of printing by a printer an image processed by said method.

60. A method according to claim 1, further comprising the step of inputting the pixel image data by a scanner.

24. An apparatus for selecting blocks of pixels from pixel image data, said apparatus under control of a computer program, said apparatus comprising:

outlining means for outlining contours of connected components in the pixel data;

forming means for forming a rectangle around each connected component outlined by said outlining means;

tree forming means for forming a hierarchical tree based on the outlined connected components, and for designating as a descendent a connected component which is within a rectangle formed around another connected component;

first connecting means for selectably connecting rectangles widthwisely based on size and proximity to other rectangles to form text lines;

second connecting means for vertically connecting a plurality of text lines formed by said first connecting means, based on size and proximity to other formed text lines, to form blocks, at least one formed text block having a plurality of formed text lines; and

modifying means for modifying the hierarchical tree based on the first and second connections.

25. An apparatus according to claim 24, further comprising input means for inputting pixel image data, said input means including converting means for converting the pixel image data into binary pixel image data in the event that the pixel image data is not binary pixel image data.

26. An apparatus according to claim 24, wherein in said forming means forms the smallest rectangle around each connected component.

27. An apparatus according to claim 24, further comprising second forming means for forming a hierarchical tree of the connected components outlined by said outlining means, based on the position of corresponding rectangles formed by said first forming means.

28. An apparatus according to claim 27, further comprising classifying means for classifying the rectangles formed by said first forming means into text and non-text units.

29. An apparatus according to claim 28, further comprising recognizing means for recognizing character images in blocks of text units.

30. An apparatus according to claim 28, wherein said classifying means classifies the rectangles in accordance with a predetermined threshold of unit height.

31. An apparatus according to claim 28, wherein said classifying means classifies the rectangles in accordance with statistical information concerning the height of said rectangles formed by said first forming means.

32. An apparatus according to claim 28, wherein said first connecting means and said second connecting means are not applied to non-text units.

33. An apparatus according to claim 28, further comprising deriving means for deriving white contours for non-text units.

34. An apparatus according to claim 33, wherein non-text units are designated as table units in accordance with the number of white contours.

35. An apparatus according to claim 33, further comprising calculating means for calculating white contour filling rate.

36. An apparatus according to claim 35, wherein the non-text unit is not designated as image data in the event that the filling rate is high.

37. An apparatus according to claim 35, further comprising recombining means for recombining non-grid-arranged white contours.

38. An apparatus according to claim 37, wherein the non-text unit is designated as a table unit in the event that a rate of recombination in said recombining means is not high.

39. An apparatus according to claim 33, wherein white contours are calculated in 4-directions.

40. An apparatus according to claim 24, wherein contours of connected components are outlined in at least 8-directions.

41. An apparatus according to claim 24, wherein said outlining means outlines the contours of connected components only at exterior portions of the connected components.

42. An apparatus according to claim 24, further comprising detecting means for detecting gaps in the pixel image data, and wherein said first connecting means does not connect rectangles into lines in the event that a gap separates the rectangles.

43. An apparatus according to claim 42, wherein columns are detected in accordance with vertically extending gaps between rectangles.

44. An apparatus according to claim 24, wherein said second connecting means includes determining means for determining non-text boundaries between lines of text data connected by said first connecting means, and wherein said second connecting means does not connect lines vertically into blocks in the event that there is an intervening non-text boundary.

45. An apparatus according to claim 24, further comprising compressing means for compressing the pixel image data before outlining.

46. Apparatus according to claim 24, further comprising means for classifying as text and non-text:

(i) hierarchical tree components; and

(ii) their descendents.

61. An apparatus according to claim 24, further comprising a printer engine for printing an image processed by said apparatus.

62. An apparatus according to claim 24, further comprising a scanner for inputting the pixel image data.

47. A computer-implemented method of selecting blocks of pixels from pixel image data comprising, in order, the steps of:

searching for connected components in the pixel data;

classifying the connected components into text components and non-text components;

forming a hierarchical tree based on the classified connected components;

forming a rectangle around each connected component and designating as a descendent a connected component which is within a rectangle formed around another connected component;

grouping text components into horizontal text lines;

grouping the grouped horizontal text lines vertically into text blocks, at least one grouped text block having a plurality of grouped horizontal text lines; and

modifying the hierarchical tree based on the grouping steps.

48. A method according to claim 47, wherein said searching step comprises the step of outlining contours of connected components and forming a rectangle around each outlined connected component.

49. A method according to claim 47, wherein said grouping steps are not performed for non-text components.

50. A method according to claim 49, further comprising the step of deriving white contours for non-text components.

51. A method according to claim 50, wherein white contours are calculated in four-directions.

52. A method according to claim 47, further comprising the step of detecting gaps in pixel image data, and wherein said grouping of horizontal lines is not performed in the event that a gap separates the horizontal lines.

53. A method according to claim 52, wherein columns are detected in accordance with vertically extending gaps.

54. A method according to claim 47, further comprising the step of detecting the direction of text lines.

55. A method according to claim 54, further comprising the step of rotating the image in the event that the direction of text lines is not horizontal.

56. A method according to claim 47, further comprising the step of searching for white gaps, and wherein in said step of grouping horizontal lines, horizontal lines are not grouped vertically into text blocks in the case where they are separated by white gaps.

57. A method according to claim 47, wherein said classifying step includes classification of text components and non-text components in accordance with size.

58. A method according to claim 57, further comprising the step of forming title lines by combining non-text components of intermediate size with adjacent text components.

63. A method according to claim 47, further comprising the step of printing by a printer an image processed by said method.

64. A method according to claim 47, further comprising the step of inputting the pixel image data by a scanner.

65. A method according to claim 47, further comprising the step of classifying as text and non text:

(i) hierarchical tree components; and

(ii) their descendents.

66. Computer-executable process steps stored on a computer readable medium, said process steps for selecting blocks of pixels from pixel image data, said process steps comprising:

an outlining step to outline contours of connected components in the pixel data;

a first forming step to form a rectangle around each connected component outlined in the outlining step;

a second forming step to form a hierarchical tree based on the outlined connected components, wherein the second forming step further comprises a designating step to designate as a descendent a connected component which is within a rectangle formed around another connected component;

a first connecting step to selectably connect rectangles widthwisely based on size and proximity to other rectangles to form text lines;

a second connecting step to selectably connect vertically a plurality of text lines formed in the first connecting step based on size and proximity to other formed text lines to form text blocks, at least one formed text block having a plurality of formed text lines; and

a modifying step to modify the hierarchical tree based on the first and second connecting steps.

67. Computer-executable process steps according to claim 66, further comprising the step of classifying as text and non-text:

(i) hierarchical tree components; and

(ii) their descendents.