Method and apparatus for pattern recognition

• US 5,524,065 A
• Filed: 07/25/1994
• Issued: 06/04/1996
• Est. Priority Date: 02/07/1992
• Status: Expired due to Term

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Yagasaki, Toshiaki
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): ANDERSON, DAVID
• Application Number:
  US08/280,248
• Time in Patent Office: 680 Days
• Field of Search: 382/13, 382/14, 382/30, 382/36, 382/38, 382/39, 382/48, 382/187, 382/186, 382/225, 382/226, 382/227, 382/155, 382/159, 382/157, 382/181, 381/42, 381/43
• US Class Current: 382/226
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches

Abstract

Pattern recognition system which provides an indication of the confidence with which a candidate is selected for an unknown pattern. The pattern recognition apparatus includes an image data input device, a host for segmenting the image data into unknown patterns, and a character recognition device for providing a candidate for each unknown pattern. The character recognition device includes a confidence level indicator for providing a confidence level indication. In one aspect, the confidence level indication is determined based on the proximity of an unknown pattern to a known pattern. In another aspect, the confidence level indication is determined based on the consistency with which the unknown pattern is recognized using different recognition functions. In yet another aspect, the confidence level indication is determined by ensuring that a candidate is not provided unless the candidate is closer than a predetermined distance from a known pattern. The pattern recognition apparatus may be provided in a stand-alone device including the image data input device, the host and the character recognition device in one integrated device, and this device may interface to a network bus. Alternatively, the pattern recognition apparatus may be distributed over a network bus. A preview function may be provided so as to preview results of recognition processing.

Claims

1. Pattern recognition apparatus which provides an identity for an unknown pattern, said apparatus comprising:

pattern recognition means for providing plural candidates for the unknown pattern, each of the plural candidates being based on a different one of respectively different distance functions;

comparing means for comparing said plural candidates so as to determine whether or not said plural candidates are comparable, and for outputting one of said plural candidates as the identity of the unknown pattern in a case where said comparing means determines that said plural candidates are comparable; and

detail processing means for performing detail processing to determine the identity of the unknown pattern, said detail processing means being selectably operable so as to determine the identity of the unknown pattern in accordance with a determination by said comparing means that said plural candidates are not comparable.

2. Apparatus according to claim 1, wherein said pattern recognition means includes classification means for providing distance values between the unknown pattern and at least one known pattern.

3. Apparatus according to claim 2, further comprising confidence level determining means for determining confidence level of said plural candidates based on the distance values from said classification means.

4. Apparatus according to claim 3, wherein said classification means selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

5. Apparatus according to claim 3, wherein said confidence level determining means determines the confidence level based on a comparison of the distance values returned by the different distance functions.

6. Apparatus according to claim 1, wherein said pattern recognition means includes classification means for providing the plural candidates in a case where the candidates are closer than a predetermined level from a known pattern.

7. Apparatus according to claim 3, wherein said detail processing means selectively performs detail processing in accordance with the confidence level.

8. Apparatus according to claim 1, wherein said unknown pattern is a character pattern.

9. Apparatus according to claim 8, further comprising discrimination means for discriminating between character patterns and non-character patterns and wherein said pattern recognition means provides a candidate in the case where a character pattern is discriminated.

10. Apparatus according to claim 8, further comprising word processing means which provides character patterns in correspondence with a word processing document, and further comprising preview means for causing character patterns provided by said word processing means to be processed by said pattern recognition means.

11. Apparatus according to claim 3, wherein said pattern recognition means provides the confidence level together with the identity of the unknown pattern.

12. Pattern recognition apparatus which provides an identity for an unknown pattern, said apparatus comprising:

a host for segmenting image information into patterns whose identities are unknown;

pattern recognition means for providing plural candidates for each unknown pattern, each of the plural candidates being based on a different one of plural respectively different distance functions;

comparing means for comparing said plural candidates so as to determine whether or not said plural candidates are comparable, and for outputting one of said plural candidates as the identity of the unknown pattern in a case where said comparing means determines that said plural candidates are comparable; and

detail processing means for performing detail processing to determine the identity of the unknown pattern, said detail processing means being selectably operable so as to determine the identity of the unknown pattern in accordance with a determination by said comparing means that said plural candidates are not comparable.

13. Apparatus according to claim 12, further comprising an image information input device for inputting said image information to the host.

14. Apparatus according to claim 12, further comprising a computerized network bus which carries image information and commands, and to which said host is connected, and wherein said host is responsive to commands on said network bus for pattern recognition.

15. Apparatus according to claim 12, wherein said pattern recognition means includes classification means for providing distance values between the unknown pattern and at least one known pattern.

16. Apparatus according to claim 15, further comprising confidence level determining means for determining the confidence level of each of said plural candidates based on the distance values from said classification means.

17. Apparatus according to claim 16, wherein said classification means selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

18. Apparatus according to claim 16, wherein said confidence level determining means determines the confidence level of said plural candidates based on a comparison of the distance values returned by the different distance functions.

19. Apparatus according to claim 12, wherein said pattern recognition means incudes classification means for providing the plural candidates in a case where the candidates are closer than a predetermined level from a known pattern.

20. Apparatus according to claim 16, wherein said detail processing means selectively performs detail processing in accordance with the confidence level.

21. Apparatus according to claim 12, wherein said unknown pattern is a character pattern.

22. Apparatus according to claim 21, further comprising discrimination means for discriminating between character patterns and non-character patterns and wherein said pattern recognition means provides a candidate in the case where a character pattern is discriminated.

23. Apparatus according to claim 21, wherein said host includes word processing means which provides character patterns in correspondence with a word processing document, and further comprising preview means for causing character patterns provided by said word processing means to be processed by said pattern recognition means.

24. Apparatus according to claim 16, wherein said host provides a reject signal in a case where the confidence level is low.

25. Pattern recognition apparatus which provides an identity for an unknown pattern, said apparatus comprising:

a computerized network bus for carrying image information;

an image input device connected to said network bus, said image input device for inputting image information;

pattern recognition means connected to said bus, said pattern recognition means for providing plural candidates for an unknown pattern, each of the plural candidates being based on a different one of plural respectively different distance functions, for comparing said plural candidates so as to determine whether or not said plural candidates are comparable, for outputting one of said plural candidates as the identity of the unknown pattern in a case where it is determined that the plural candidates are comparable, and for performing detail processing to determine the identity of the unknown pattern in a case where it is determined that the plural candidates are not comparable; and

plural processing means connected to said bus, each said processing means for segmenting image information input by said image input device into patterns whose identities are unknown, for providing each unknown pattern to said pattern recognition means, and for receiving the identity of the unknown pattern determined from said pattern recognition means.

26. Apparatus according to claim 25, wherein said pattern recognition means includes classification means for providing distance values between the unknown pattern and at least one known pattern.

27. Apparatus according to claim 26, wherein said pattern recognition means further comprises confidence level determining means for determining confidence level of said plural candidates based on the distance values from said classification means.

28. Apparatus according to claim 27, wherein said pattern recognition means provides the confidence level together with the identity of the unknown pattern.

29. Apparatus according to claim 27, wherein said classification means selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

30. Apparatus according to claim 27, wherein said confidence level determining means determines the confidence level based on a comparison of the distance values returned by the different distance functions.

31. Apparatus according to claim 25, wherein said pattern recognition means includes classification means for providing the candidate in a case where the candidate is closer than a predetermined level from a known pattern.

32. Apparatus according to claim 25, wherein said unknown pattern is a character pattern.

33. Apparatus according to claim 32, further comprising discrimination means for discriminating between character patterns and non-character patterns and wherein said pattern recognition means provides a candidate in the case where a character pattern is discriminated.

34. Apparatus according to claim 32, wherein each of said plural processing means includes word processing means which provides character patterns in correspondence with a word processing document, and further comprising preview means for causing character patterns provided by said word processing means to be processed by said pattern recognition means.

35. A pattern recognition method for providing an identity for an unknown pattern, comprising the steps of:

determining distance values between an unknown pattern and at least one known pattern based on plural respectively different distance functions;

providing plural candidates for the unknown pattern, each of the plural candidates corresponding to a different one of the plural respectively different distance functions;

comparing said plural candidates so as to determine whether or not said plural candidates are comparable;

outputting one of said plural candidates as the identity of the unknown pattern in a case where said comparing step determines that said plural candidates are comparable; and

performing detail processing to determine the identity of the unknown pattern, said detail processing step being selectively performed so as to determine the identity of the unknown pattern in a case where said comparing step determines that said plural candidates are not comparable.

36. A method according to claim 35, wherein said step of determining distance values includes calculating the distance values between the unknown pattern and the at least one known pattern.

37. A method according to claim 36, further comprising confidence level determining step for determining confidence level of said plural candidates based on distance values provided by said different distance functions.

38. A method according to claim 37, wherein said providing step selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

39. A method according to claim 37, wherein said confidence level determining step determines the confidence level based on a comparison of the distance values returned by the different distance functions.

40. A method according to claim 35, wherein said step of providing the plural candidates includes providing the plural candidates in a case where the plural candidates are closer than a predetermined level from a known pattern.

41. A method according to claim 35, wherein said unknown pattern is a character pattern.

42. A method according to claim 41, further comprising the step of discriminating between character patterns and non-character patterns and wherein said providing step provides a candidate in the case where a character pattern is discriminated.

43. A method according to claim 41, further comprising the steps of processing a document file so as to obtain character patterns, providing a candidate for each character pattern so obtained, and previewing the candidates with respect to the document file.

44. A method according to claim 37, further comprising the step of providing a reject signal in a case where the confidence level is low.

45. A pattern recognition method for providing an identity for an unknown pattern segment, comprising the steps of:

segmenting image information into unknown pattern segments;

determining distance values between an unknown pattern segment and at least one known pattern segment based on plural respectively different distance functions;

providing plural candidates for the unknown pattern segment, each of the plural candidates corresponding to a different one of the plural respectively different distance functions;

comparing said plural candidates so as to determine whether or not said plural candidates are comparable;

outputting one of said plural candidates as the identity of the unknown pattern segment in a case where said comparing step determines that said plural candidates are comparable; and

performing detail processing to determine the identity of the unknown pattern segment, wherein said detail processing step is selectably performed so as to determine the identity of the unknown pattern segment in a case where said comparing step determines that said plural candidates are not comparable.

46. A method according to claim 45, further comprising the step of inputting said image information to a host.

47. A method according to claim 45, further comprising the step of connecting said host to a computerized network bus which carries image information and commands, wherein said host is responsive to commands on said network bus for pattern recognition.

48. A method according to claim 45, wherein said step of determining distance values includes calculating the distance values between the unknown pattern and the at least one known pattern.

49. A method according to claim 48, further comprising a confidence level determining step determining step for determining confidence level of said plural candidates based on the distance values provided by the different distance functions.

50. A method according to claim 49, wherein said providing step selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

51. A method according to claim 49, wherein said confidence level determining step determines the confidence level based on a comparison of the distance values returned by the different distance functions.

52. A method according to claim 45, wherein said candidate providing step includes providing the candidate in a case where the candidate is closer than a predetermined level from a known pattern.

53. A method according to claim 49, wherein said detail processing step is selectively performed in accordance with the confidence level.

54. A method according to claim 45, wherein said unknown pattern is a character pattern.

55. A method according to claim 54, further comprising the step of discriminating between character patterns and non-character patterns and wherein said candidate providing step provides a candidate in the case where a character pattern is discriminated.

56. A method according to claim 54, further comprising the steps of processing a document file so as to obtain character patterns, providing a candidate for each character pattern so obtained, and previewing the candidates with respect to the document file.

57. A method according to claim 49, further comprising the step of providing a reject signal in a case where the confidence level is low.

58. A pattern recognition method for providing an identity for an unknown pattern segment, comprising the steps of:

inputting image information from an image input device connected to a computerized network bus which carries such image information;

segmenting said image information into segmented patterns whose identities are unknown;

determining distance values between an unknown segmented pattern and at least one known pattern based on plural respectively different distance functions;

providing plural candidates for each of the unknown segmented patterns, each of the plural candidates being based on a different one of the plural respectively different distance functions;

comparing said plural candidates so as to determine whether or not said plural candidates are comparable;

selecting one of said plural candidates as the identity of the unknown pattern segment in a case where said comparing step determines that said plural candidates are comparable;

performing detail processing to determine the identity of the unknown pattern segment, said detail processing step being selectively performed so as to determine the identity of the unknown pattern segment in a case where said comparing step determines that said plural candidates are not comparable; and

transferring the identity for each unknown pattern segment onto the network bus.

59. A method according to claim 58, further comprising a confidence level determining step for determining the confidence level of said plural candidates based on the distance values provided by said different distance functions.

60. A method according to claim 59, further comprising the step of processing the identified candidate for each unknown pattern segment transferred onto the network bus and providing the identified candidate for each unknown pattern segment in the case where the confidence level is high.

61. A method according to claim 58, wherein said step of determining distance values includes calculating the distance values between the segmented pattern and the at least one known pattern.

62. A method according to claim 59, wherein said providing step selectably provides sorted distance values between the unknown pattern and the at least one known pattern, and sorted candidates for the unknown pattern.

63. A method according to claim 59, wherein said confidence level determining step determines the confidence level based on a comparison of the distance values returned by the different distance functions.

64. A method according to claim 58, wherein said candidate providing step includes providing the candidate in a case where the candidate is closer than a predetermined level from a known pattern.

65. A method according to claim 58, wherein said unknown pattern is a character pattern.

66. A method according to claim 65, further comprising the step of discriminating between character patterns and non-character patterns and wherein said candidate providing step provides a candidate in the case where a character pattern is discriminated.

67. A method according to claim 65, further comprising the steps of processing a document file so as to obtain character patterns, providing a candidate for each character pattern so obtained, and previewing the candidates with respect to the document file.

68. A method according to claim 59, further comprising the step of providing a reject signal in a case where the confidence level is low.