MATCHING POINT EXTRACTING METHOD AND APPARATUS THEREFOR

• US 20020039438A1
• Filed: 02/20/1997
• Published: 04/04/2002
• Est. Priority Date: 02/21/1996
• Status: Active Grant

• Current Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Original Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Inventors: YANO, KOTARO, MORI, KATSUHIKO, IIJIMA, KATSUMI
• Granted Patent: US 6,493,465 B2
• Time in Patent Office: Days
• US Class Current: 382/154
• CPC Class Codes: G06T 7/32 using correlation-based methods

Abstract

A thermoplastic resin composition comprising a thermoplastic resin (A) not forming an anisotropic melt phase and a liquid crystal polymer (B) capable of forming an anisotropic melt phase, an injection molding method thereof and an injection molded article capable of providing a thin molded article having excellent mechanical strength.

A thermoplastic resin composition (C) comprises 99 to 50 wt. % of the thermoplastic resin (A) and 1 to 50 wt. % of the liquid crystal polymer (B). The liquid crystal polymer (B) is microscopically dispersed in an island form in the matrix phase of the thermoplastic resin (A) so that when the composition (C) is heat-treated at a temperature not lower than the melting point of the liquid crystal polymer (B) under a non-load state and is then cooled, the liquid crystal polymer (B) has a weight average particle diameter within the range of 10 to 40 micrometers and at least 80 wt. % thereof has a particle diameter within the range of 0.5 to 60 micrometers. The injection molding method comprises carrying out injection molding of the thermoplastic resin composition (C) at a processing temperature greater than or equal to the flow beginning temperature of the liquid crystal polymer (B) and at a gate passage rate of the molten resin of at least 500 m/min, and an injection molded article characterized in that the liquid crystal polymer (B) is dispersed in a fibrous form in the matrix phase of the thermoplastic resin (A) at a mean aspect ratio of at least 6.

Claims

1. A matching point extracting apparatus for extracting, in synthesizing plural images, the matching points of said plural images, comprising:

A) template image data formation means for entering plural image data each corresponding to an image of an image frame, taking an image represented by the image data corresponding to an arbitrary image frame among thus entered image data of plural image frames as a reference image, extracting the image data, corresponding to a partial area in said reference image, as a template image and outputting the template image data corresponding to thus extracted template image;

B) matching candidate extraction means for comparing the value of each pixel data constituting the template image data outputted by said template image data formation means with the value of each pixel data constituting the image data corresponding to each of search images which are the images, other than the reference image, of the image data of the plural image frames entered into said template image data formation means, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said search image to said template image according to said number of pixel data, extracting said search image showing maximum in said similarity as a matching candidate for said template image, and outputting said matching candidate together with similarity data indicating said similarity;

C) decision means for judging, based on the similarity data outputted by said matching candidate extraction means, whether the pixel data constituting said template image satisfy a specified condition; and

D) template image changing means adapted, in case said decision means identifies that the pixel data constituting said template image satisfy said specified condition, for changing the shape of the template image represented by the template image data outputted by said template image formation means and outputting such changed template image to said matching candidate extraction means.

2. A matching point extracting apparatus according to claim 1, wherein said matching candidate extraction means is adapted to use, as the value of said pixel data, the value of the luminance component.

3. A matching point extracting apparatus according to claim 1, wherein said matching candidate extraction means is adapted to use, as the value of said pixel data, the value of a color component.

4. A matching point extracting apparatus according to claim 2, wherein said decision means is so arranged as to judge whether each pixel data constituting the template image outputted by said matching candidate extraction means is a pixel data in an edge area in which the distribution of the values of the luminance component shows a steep variation.

5. A matching point extracting apparatus according to claim 3, wherein said decision means is so arranged as to judge whether each pixel data constituting the template image outputted by said matching candidate extraction means is a pixel data in an edge area in which the distribution of the values of the color component shows a steep variation.

6. A matching point extracting apparatus according to claim 4, wherein said template image changing means is so arranged as, in case said decision means identifies that each pixel data constituting said template image satisfies said specified condition, to reduce the area, by a predetermined amount, of the template image represented by the template image data outputted by said template image data formation means.

7. A matching point extracting apparatus according to claim 4, wherein said template image changing means is so arranged as, in case said decision means identifies that each pixel data constituting said template image satisfies said specified condition, to mask an area not containing the center of said template image, among areas divided by a strong edge in the template image represented by the template image data outputted by said template image data formation means.

8. A matching point extracting apparatus for extracting, in synthesizing plural images, the matching points of said plural images, comprising:

A) template image data formation means for entering plural image data each corresponding to an image of an image frame, taking an image represented by the image data corresponding to an arbitrary image frame among thus entered image data of plural image frames as a reference image, extracting the image data, corresponding to a partial area in said reference image, as a template image and outputting the template image data corresponding to thus extracted template image;

B) first matching candidate extraction means for comparing the value of each pixel data constituting the template image data outputted by said template image data formation means with the value of each pixel data constituting the image data corresponding to each of search images which are the images, other than the reference image, of the image data of the plural image frames entered into said template image data formation means, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said search image to said template image according to said number of pixel data, extracting said search image showing maximum in said similarity as a matching candidate for said template image, and outputting said matching candidate together with similarity data indicating said similarity;

C) template image data formation instruction means adapted, based on the matching candidate extracted by said first matching candidate extraction means, to instruct said template image data formation means to form template image data from said search image;

D) second matching candidate extraction means adapted, in response to the instruction by said template image data formation instruction means, for comparing the value of each pixel data constituting the template image data formed from said search image in said template image data formation means with the value of each pixel data constituting the image data corresponding to said reference image, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said reference image to said template image formed from said search image according to said number of pixel data, extracting a pixel area showing maximum in said similarity from said reference image as a matching candidate for said template image formed from the search image, and outputting said matching candidate together with similarity data indicating said similarity; and

E) decision means adapted, in case the matching candidate extracted from said reference image by said second matching candidate extraction means coincides with the matching candidate extracted from said search image by said first matching candidate extraction means, in extracting the respective matching candidates as the matching points, for judging, based on said similarity data, whether each pixel data constituting said template image is a pixel data in a pixel area satisfying specified characteristics.

9. A matching point extracting apparatus according to claim 8, wherein said decision means is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an area showing a steep variation in the parallax distribution.

10. A matching point extracting apparatus according to claim 8, wherein said decision means is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an occlusion area which has no matching candidate.

11. A matching point extracting apparatus according to claim 10, wherein said first and second matching candidate extraction means are adapted to use, as the value of said pixel data, the value of the luminance component.

12. A matching point extracting apparatus according to claim 10, wherein said first and second matching candidate extraction means are adapted to use, as the value of said pixel data, the value of a color component.

13. A matching point extracting apparatus according to claim 8, further comprising control means adapted, in case said decision means identifies that each pixel data constituting said template image is a pixel data in a pixel area satisfying specified characteristics, for controlling said template image data formation means and said template image data formation instruction means so as to change the shapes of the template images formed from said reference image and said search image.

14. A matching point extracting apparatus according to claim 13, wherein said decision means is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an area showing a steep variation in the parallax distribution.

15. A matching point extracting apparatus according to claim 14, wherein said control means is adapted, in case said decision means identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an area showing a steep variation in the parallax distribution, for controlling said template image data formation means and said template image data formation instruction means so as to reduce the size of the template images formed from said reference image and said search image.

16. A matching point extracting apparatus according to claim 14, wherein said control means is adapted, in case said decision means identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an area showing a steep variation in the parallax distribution, for controlling said template image data formation means and said template image data formation instruction means so as to mask an area not containing the center of said template image, among areas divided by a strong edge in the template images formed from said reference image and said template image.

17. A matching point extracting apparatus according to claim 13, wherein said decision means is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an occlusion area which has no matching candidate.

18. A matching point extracting apparatus according to claim 17, wherein said control means is adapted, in case said decision means identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an occlusion area having no matching candidate, for controlling said template image data formation means and said template image data formation instruction means so as to reduce the size of the template images formed from said reference image and said search image.

19. A matching point extracting apparatus according to claim 17, wherein said control means is adapted, in case said decision means identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction means is a pixel data in an occlusion area having no matching candidate, for controlling said template image data formation means and said template image data formation instruction means so as to mask an area not containing the center of said template image, among areas divided by a strong edge in the template images formed from said reference image and said template image.

20. A matching point extracting method for extracting, in synthesizing plural images, the matching points of said plural images, comprising:

A) a template image data formation step of entering plural image data each corresponding to an image of an image frame, taking an image represented by the image data corresponding to an arbitrary image frame among thus entered image data of plural image frames as a reference image, extracting the image data, corresponding to a partial area in said reference image, as a template image and outputting the template image data corresponding to thus extracted template image;

B) a matching candidate extraction step of comparing the value of each pixel data constituting the template image data outputted by said template image data formation step with the value of each pixel data constituting the image data corresponding to each of search images which are the images, other than the reference image, of the image data of the plural image frames entered in said template image data formation step, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said search image to said template image according to said number of pixel data, extracting said search image showing maximum in said similarity as a matching candidate for said template image, and outputting said matching candidate together with similarity data indicating said similarity;

C) a decision step of judging, based on the similarity data outputted by said matching candidate extraction step, whether the pixel data constituting said template image satisfy a specified condition; and

D) a template image changing step adapted, in case said decision step identifies that the pixel data constituting said template image satisfy said specified condition, for changing the shape of the template image represented by the template image data outputted by said template image represented by the template image data outputted by said template image formation step and outputting such changed template image to said matching candidate extraction step.

21. A matching point extracting method according to claim 20, wherein said matching candidate extraction step is adapted to use, as the value of said pixel data, the value of the luminance component.

22. A matching point extracting method according to claim 20, wherein said matching candidate extraction step is adapted to use, as the value of said pixel data, the value of a color component.

23. A matching point extracting method according to claim 21, wherein said decision step is adapted to judge whether each pixel data constituting the template image outputted by said matching candidate extraction step is a pixel data in an edge area in which the distribution of the values of the luminance component shows a steep variation.

24. A matching point extracting method according to claim 22, wherein said decision step is adapted to judge whether each pixel data constituting the template image outputted by said matching candidate extraction step is a pixel data in an edge area in which the distribution of the values of the color component shows a steep variation.

25. A matching point extracting method according to claim 23, wherein said template image changing step is adapted, in case said decision step identifies that each pixel data constituting said template image satisfies said specified condition, to reduce the area, by a predetermined amount, of the template image represented by the template image data outputted by said template image data formation step.

26. A matching point extracting method according to claim 23, wherein said template image changing step is adapted, in case said decision step identifies that each pixel data constituting said template image satisfies said specified condition, to mask an area not containing the center of said template image, among areas divided by a strong edge in the template image represented by the template image data outputted by said template image data formation step.

28. A matching point extracting method according to claim 27, wherein said decision step is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an area showing a steep variation in the parallax distribution.

29. A matching point extracting method according to claim 27, wherein said decision step is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an occlusion area which has no matching candidate.

30. A matching point extracting method according to claim 29, wherein said first and second matching candidate extraction steps are adapted to use, as the value of said pixel data, the value of the luminance component.

31. A matching point extracting method according to claim 29, wherein said first and second matching candidate extraction step are adapted to use, as the value of said pixel data, the value of a color component.

32. A matching point extracting method according to claim 27, further comprising a control step adapted, in case said decision step identifies that each pixel data constituting said template image is a pixel data in a pixel area satisfying specified characteristics, for controlling said template image data formation step and said template image data formation instruction step so as to change the shapes of the template images formed from said reference image and said search image.

33. A matching point extracting method according to claim 27, wherein said decision means is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an area showing a steep variation in the parallax distribution.

34. A matching point extracting method according to claim 33, wherein said control step is adapted, in case said decision step identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an area showing a steep variation in the parallax distribution, for controlling said template image data formation step and said template image data formation instruction step so as to reduce the size of the template images formed from said reference image and said search image.

35. A matching point extracting method according to claim 33, wherein said control step is adapted, in case said decision step identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an area showing a steep variation in the parallax distribution, for controlling said template image data formation step and said template image data formation instruction step so as to mask an area not containing the center of said template image, among areas divided by a strong edge in the template images formed from said reference image and said template image.

36. A matching point extracting method according to claim 27, wherein said decision step is adapted to judge whether each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an occlusion area which has no matching candidate.

37. A matching point extracting method according to claim 36, wherein said control step is adapted, in case said decision step identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an occlusion area having no matching candidate, for controlling said template image data formation step and said template image data formation instruction step so as to reduce the size of the template images formed from said reference image and said search image.

38. A matching point extracting method according to claim 36, wherein said control step is adapted, in case said decision step identifies that each pixel data constituting the matching candidate extracted by said second matching candidate extraction step is a pixel data in an occlusion area having no matching candidate, for controlling said template image data formation step and said template image data formation instruction step so as to mask an area not containing the center of said template image, among areas divided by a strong edge in the template images formed from said reference image and said template image.

40. A matching point extracting method according to claim 39, wherein said plural templates are mutually different in the areas thereof.

41. A matching point extracting method according to claim 39, wherein said plural templates are mutually different in the shapes thereof, according to the edge distribution therein.

42. A matching point extracting method according to claim 39, wherein said matching point decision step is adapted to determine the matching points among the plural image by employing the plural templates in succession, and the operation executed with the second or subsequent template is executed in an area including a candidate matching point detected in a preceding operation.

43. A matching point extracting method according to claim 39, wherein said matching point decision step is adapted to determine the matching points among the plural image by employing the plural templates in succession, and the operation with the second or subsequent template is executed after the judgment of the result of a preceding operation.

44. A matching point extracting method according to claim 39, wherein said matching point decision step is adapted to determine the matching points among the plural image by simultaneously employing the plural templates, and to select the result of an operation among the results of such plural operations.

45. A matching point extracting method according to claim 39, wherein, in said matching point decision step, the image area showing a specified condition is an image area involving steep variation in the parallax distribution among the plural images.

46. A matching point extracting method according to claim 45, wherein, in said matching point decision step, the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images is executed by the evaluation value in the template matching method.

47. A matching point extracting method according to claim 45, wherein, in said matching point decision step, the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images is executed by the evaluation value in the template matching method and the result of judgment whether a reference point coincides with a candidate matching point of said reference point.

48. A matching point extracting method according to claim 45, wherein, in said matching point decision step, in the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images, the number of pixels showing differences in the pixels values smaller than a certain threshold value between the pixels within the template and a pixel in a searched image is used as the evaluation function for the similarity in the template matching method.

50. A matching point extracting apparatus according to claim 49, wherein said plural templates are mutually different in the areas thereof.

51. A matching point extracting apparatus according to claim 49, wherein said plural templates are mutually different in the shapes thereof, according to the edge distribution therein.

52. A matching point extracting apparatus according to claim 49, wherein said matching point decision means is adapted to determine the matching points among the plural image by employing the plural templates in succession, and the operation executed with the second or subsequent template is executed in an area including a candidate matching point detected in a preceding operation.

53. A matching point extracting apparatus according to claim 49, wherein said matching point decision means is adapted to determine the matching points among the plural image by employing the plural templates in succession, and the operation with the second or subsequent template is executed after the judgment of the result of a preceding operation.

54. A matching point extracting apparatus according to claim 49, wherein said matching point decision means is adapted to determine the matching points among the plural image by simultaneously employing the plural templates, and to select the result of an operation among the results of such plural operations.

55. A matching point extracting apparatus according to claim 49, wherein, in said matching point decision means, the image area showing a specified condition is an image area involving steep variation in the parallax distribution among the plural images.

56. A matching point extracting apparatus according to claim 55, wherein, in said matching point decision means, the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images is executed by the evaluation value in the template matching method.

57. A matching point extracting apparatus according to claim 55, wherein, in said matching point decision means, the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images is executed by the evaluation value in the template matching method and the result of judgment whether a reference point coincides with a candidate matching point of said reference point.

58. A matching point extracting apparatus according to claim 55, wherein, in said matching point decision means, in the judgment whether the image area is an image area involving steep variation in the parallax distribution among the plural images, the number of pixels showing differences in the pixels values smaller than a certain threshold value between the pixels within the template and a pixel in a searched image is used as the evaluation function for the similarity in the template matching method.

60. A matching point extracting method according to claim 59, wherein said second template formation step is adapted to form a template which is different in area from the template formed in said first template formation step.

61. A matching point extracting method according to claim 59, wherein, in said decision step, the judgment of the specified condition is achieved by comparison of the similarity determined in said similarity operation step and a certain threshold value.

62. A matching point extracting method according to claim 59, wherein, in said decision step, the specified condition is judged as satisfied in case the similarity determined in said similarity operation step is smaller than a certain threshold value.

63. A matching point extracting method according to claim 62, wherein said second template formation step forms a template different in area from the template formed in said first template formation step, based on the result of judgment in said decision step.

64. A matching point extracting method according to claim 59, wherein, in said decision step, the specified condition is judged as satisfied in case the difference of plural similarities determined in said similarity operation step is smaller than a certain threshold value.

65. A matching point extracting method according to claim 64, wherein said second template formation step forms a template different in area from the template formed in said first template formation step, based on the result of judgment in said decision step.

66. A matching point extracting method according to claim 59, wherein the specified condition in said decision step is whether the image area is an image area involving steep variation in the parallax distribution among said plural images.

67. A matching point extracting method according to claim 66, wherein said second template formation step forms a template different in area from the template formed in said first template formation step, based on the result of judgment in said decision step.

68. A matching point extracting method according to claim 59, wherein, in said first template formation step, said predetermined judgment standard for forming the initial template is whether a strong edge is present in an image area indicated by the template.

70. A matching point extracting apparatus according to claim 69, wherein said second template formation means is adapted to form a template which is different in area from the template formed by said first template formation means.

71. A matching point extracting apparatus according to claim 69, wherein, in said decision means, the judgment of the specified condition is achieved by comparison of the similarity determined by said similarity operation means and a certain threshold value.

72. A matching point extracting apparatus according to claim 69, wherein, in said decision means, the specified condition is judged as satisfied in case the similarity determined in said similarity operation means is smaller than a certain threshold value.

73. A matching point extracting apparatus according to claim 72, wherein said second template formation means forms a template different in area from the template formed by said first template formation means, based on the result of judgment by said decision means.

74. A matching point extracting apparatus according to claim 69, wherein, in said decision means, the specified condition is judged as satisfied in case the difference of plural similarities determined by said similarity operation means is smaller than a certain threshold value.

75. A matching point extracting apparatus according to claim 74, wherein said second template formation means forms a template different in area from the template formed by said first template formation means, based on the result of judgment by said decision means.

76. A matching point extracting apparatus according to claim 69, wherein the specified condition in said decision step is whether the image area is an image area involving steep variation in the parallax distribution among said plural images.

77. A matching point extracting apparatus according to claim 76, wherein said second template formation means forms a template different in area from the template formed by said first template formation means, based on the result of judgment by said decision means.

78. A matching point extracting apparatus according to claim 69, wherein, in said first template formation means, said predetermined judgment standard for forming the initial template is whether a strong edge is present in an image area indicated by the template.

27. A matching point extracting method for extracting, in synthesizing plural images, the matching points of said plural images, comprising:

A) a template image data formation step for entering plural image data each corresponding to an image of an image frame, taking an image represented by the image data corresponding to an arbitrary image frame among thus entered image data of plural image frames as a reference image, extracting the image data, corresponding to a partial area in said reference image, as a template image and outputting the template image data corresponding to thus extracted template image;

B) a first matching candidate extraction step for comparing the value of each pixel data constituting the template image data outputted by said template image data formation step with the value of each pixel data constituting the image data corresponding to each of search images which are the images, other than the reference image, of the image data of the plural image frames entered in said template image data formation step, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said search image to said template image according to said number of pixel data, extracting said search image showing maximum in said similarity as a matching candidate for said template image, and outputting said matching candidate together with similarity data indicating said similarity;

C) a template image data formation instruction step adapted, based on the matching candidate extracted by said first matching candidate extraction step, to instruct said template image data formation step to form template image data from said search image.

D) a second matching candidate extraction step adapted, in response to the instruction by said template image data formation instruction step, for comparing the value of each pixel data constituting the template image data formed from said search image in said template image data formation step with the value of each pixel data constituting the image data corresponding to said reference image, determining the number of pixel data showing differences smaller than a predetermined value, calculating the similarity of said reference image to said template image formed from said search image according to said number of pixel data, extracting a pixel area showing maximum in said similarity from said reference image as a matching candidate for said template image formed from the search image, and outputting said matching candidate together with similarity data indicating said similarity; and

E) a decision step adapted, in case the matching candidate extracted from said reference image by said second matching candidate extraction step coincides with the matching candidate extracted from said search image by said first matching candidate extraction step, in extracting the respective matching candidates as the matching points, for judging, based on said similarity data, whether each pixel data constituting said template image is a pixel data in a pixel area satisfying specified characteristics.

39. A matching point extracting method for determining the matching points among plural images based on a template matching method, comprising:

A) a template formation step of forming mutually different plural templates from an arbitrary image in the plural images; and

B) a matching point decision step of determining the matching points among the plural images, in an image area showing a specified condition, by employing one of the plural templates formed in said template formation step.

49. A matching point extracting apparatus for determining the matching points among plural images based on a template matching method, comprising:

A) template formation means for forming mutually different plural templates from an arbitrary image in the plural images; and

B) matching point decision means for determining the matching points among the plural images, in an image area showing a specified condition, by employing one of the plural templates formed by said template formation means.

59. A matching point extracting method for determining the matching points among plural images based on a template matching method, comprising:

A) a first template formation step of forming an initial template, according to a predetermined judgment standard, from an arbitrary image in plural input images;

B) a similarity operation step of determining the similarity, utilizing the pixel values of the pixels of the template formed in said first template formation step and pixel value of each pixel in another of said input images;

C) a decision step of judging a specified condition from the similarity determined in said similarity operation step; and

D) a second template formation step of forming another template based on the result of judgment in said decision step.

69. A matching point extracting apparatus for determining the matching points among plural images based on a template matching method, comprising:

A) first template formation means for forming an initial template, according to a predetermined judgment standard, from an arbitrary image in plural input images;

B) similarity operation means for determining the similarity, utilizing the pixel values of the pixels of the template formed by said first template formation means and pixel value of each pixel in another of said input images;

C) decision means for judging a specified condition from the similarity determined in said similarity operation means; and

D) second template formation means for forming another template based on the result of judgment by said decision means.