Transform artifact reduction process

• US 5,852,475 A
• Filed: 06/06/1995
• Issued: 12/22/1998
• Est. Priority Date: 06/06/1995
• Status: Expired due to Term

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Compression Labs Incorporated (Asure Software)
• Inventors: Tse, Yi Tong, Gupta, Smita
• Primary Examiner(s): Grant, Chris
• Application Number:
  US08/486,279
• Time in Patent Office: 1,295 Days
• Field of Search: 348/606, 348/625, 348/607, 348/420, 348/384, 348/390, 348/699, 348/700
• US Class Current: 348/606
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: H04N 19/527 Global motion vector estimation
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness

Abstract

A post-processor for a decoded video sequence includes a digital noise reduction unit and an artifact reduction unit which significantly reduce blocking artifacts and mosquito noise in a video image. The post-processor uses both temporal and edge characteristics of the video image to enhance the displayed image. The post-processor operates on a current frame of pixel data using information from the immediately preceding post-processed frame that is stored in a frame memory of the post-processor. The post-processor first identifies texture and fine detail areas in the image. The post-processor uses artifact reduction only on portions of the image that are not part of an edge, and are not part of a texture or fine detail area. Since artifact reduction is not utilized on these areas, the post-processed image is not softened in regions where it is easily noticed by the human eye.

Claims

1. A transform artifact reduction method for decoded video pixel data, said method comprising:

digitally noise reducing a block of pixels to obtain a digitally noise reduced block of pixels; and

filtering said digitally noise reduced block of pixels using a spatially-variant filter wherein said filtering further comprises

adaptively identifying each pixel in said digitally noise reduced block of pixels as one of an edge pixel and a non-edge pixel;

assigning a pixel in said digitally noise reduced block of pixels one classification in a plurality of edge classifications using said edge pixel and said non-edge pixel identifications for pixels in a window about said pixel, said assigning said pixel in said digitally noise reduced block of pixels one classification in said plurality of edge classifications using said edge pixel and said non-edge pixel classifications for pixels in a window about said pixel further comprising

generating an edge map for said digitally noise reduced block of pixels wherein in said edge map, an edge flag for each pixel in said digitally noise reduced block of pixels is configured to identify each pixel as one of said edge pixel and said non-edge pixel; and

adaptively filtering each pixel in said digitally noise reduced block of pixels based upon said assigned classification to reduce transform artifacts in a video image.

2. A transform artifact reduction method for decoded video pixel data as in claim 1 wherein said generating said edge map further comprises:

comparing a pixel gradient for a pixel with a threshold; and

setting said edge flag for said pixel in said edge map upon said pixel gradient being greater than said threshold.

3. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said comparing a pixel gradient for said pixel further comprises:

comparing said pixel gradient with an adaptive edge threshold.

4. A transform artifact reduction method for decoded video pixel data as in claim 3 wherein said adaptive edge threshold is an adaptive texture threshold.

5. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said comparing said pixel gradient for said pixel further comprises:

comparing said pixel gradient with a luminance threshold.

6. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said comparing said pixel gradient for said pixel further comprises:

comparing said pixel gradient with both an adaptive edge threshold, and a luminance threshold.

7. A transform artifact reduction method for decoded video pixel data as in claim 6 wherein said setting said edge flag for said pixel in said edge map upon said pixel gradient being greater than said threshold further comprises:

setting said edge flag only upon said pixel gradient being greater than said adaptive edge threshold, and being greater than said luminance threshold.

8. A transform artifact reduction method for decoded video pixel data as in claim 6 further comprising prior to said comparing said pixel gradient:

generating said luminance threshold.

9. A transform artifact reduction method for decoded video pixel data as in claim 8 wherein said generating said luminance threshold further comprises:

generating a background luminance measure for said pixel wherein said luminance threshold is proportional to said background luminance measure.

10. A transform artifact reduction method for decoded video pixel data as in claim 9 wherein said generating said background luminance measure for said pixel further comprises:

averaging luminance components in a window about said pixel.

11. A transform artifact reduction method for decoded video pixel data as in claim 10 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

12. A transform artifact reduction method for decoded video pixel data as in claim 9 wherein said generating said luminance threshold further comprises:

adjusting said background luminance measure to obtain an adjusted background luminance measure.

13. A transform artifact reduction method for decoded video pixel data as in claim 12 wherein said generating said luminance threshold further comprises:

defining said luminance threshold as the maximum of a minimum luminance threshold and said adjusted luminance background measure.

14. transform artifact reduction method for decoded video pixel data as in claim 3 further comprising prior to said comparing said pixel gradient with said adaptive edge threshold:

generating said pixel gradient for said pixel.

15. A transform artifact reduction method for decoded video pixel data as in claim 14 wherein said generating said pixel gradient further comprises:

generating a plurality of spatial gradients for said pixel using pixels in a window about said pixel.

16. A transform artifact reduction method for decoded video pixel data as in claim 15 wherein said window is a three-by three pixel window with said pixel centered in said three-by-three pixel window.

17. A transform artifact reduction method for decoded video pixel data as in claim 15 wherein said plurality of spatial gradients comprise two spatial gradients.

18. A transform artifact reduction method for decoded video pixel data as in claim 15 wherein said plurality of spatial gradients comprise four spatial gradients.

19. A transform artifact reduction method for decoded video pixel data as in claim 15 wherein said generating said pixel gradient further comprises:

combining said plurality of spatial gradients to obtain said pixel gradient.

20. A transform artifact reduction method for decoded video pixel data as in claim 3 further comprising prior to said comparing said pixel gradient with said adaptive edge threshold:

generating said adaptive edge threshold.

21. A transform artifact reduction method for decoded video pixel data as in claim 20 wherein said generating said adaptive edge threshold further comprises:

generating an edge estimator for a plurality of edges through said pixel to obtain a plurality of edge estimators wherein pixels in a window about said pixel are used to generate said plurality of edge estimators.

22. A transform artifact reduction method for decoded video pixel data as in claim 21 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

23. A transform artifact reduction method for decoded video pixel data as in claim 21 wherein said plurality of edge estimators comprise two pixel texture estimators.

24. A transform artifact reduction method for decoded video pixel data as in claim 21 wherein said plurality of edge estimators comprise four pixel texture estimators.

25. A transform artifact reduction method for decoded video pixel data as in claim 21 wherein said generating said adaptive edge threshold further comprises:

combining said plurality of edge estimators to obtain said adaptive edge threshold.

26. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said plurality of edge classifications includes an edge classification, and said pixel is assigned the edge classification if said edge flag for said pixel in said edge map indicates that said pixel is an edge pixel.

27. A transform artifact reduction method for decoded video pixel data as in claim 26 wherein a pixel assigned the edge classification is unchanged by said filtering of said pixel.

28. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said plurality of edge classifications includes an edge border classification.

29. A transform artifact reduction method for decoded video pixel data as in claim 28 wherein said pixel is assigned said edge border classification if (i) the edge flag for at least three pixels in a window of pixels about said pixel are not set;

and (ii) said pixel is in a line of pixels in said window and the edge flag for each pixel in the line is not set.

30. A transform artifact reduction method for decoded video pixel data as in claim 29 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

31. A transform artifact reduction method for decoded video pixel data as in claim 29 wherein a pixel assigned said edge border classification is filtered in a one-dimensional filter that processes said line of pixels.

32. A transform artifact reduction method for decoded video pixel data as in claim 31 where said one-dimensional filter is a low pass filter.

33. A transform artifact reduction method for decoded video pixel data as in claim 2 wherein said plurality of edge classifications includes a shade classification.

34. A transform artifact reduction method for decoded video pixel data as in claim 33 wherein said pixel is assigned said shade classification if the edge flag for said pixel and each pixel in a window of pixels about said pixel is not set.

35. A transform artifact reduction method for decoded video pixel data as in claim 34 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

36. A transform artifact reduction method for decoded video pixel data as in claim 34 wherein a pixel assigned said shade classification is filtered in a two-dimensional filter that processes said window of pixels.

37. A transform artifact reduction method for decoded video pixel data as in claim 36 where said two-dimensional filter is a low pass filter.

38. A transform artifact reduction method for decoded video pixel data, said method comprising:

digitally noise reducing a block of pixels to obtain a digitally noise reduced block of pixels;

setting a replenished block flag to a predetermined state upon determining that said digitally noise reduced block of pixels is a replenished block;

identifying each pixel in said digitally noise reduced block of pixels as one of an edge pixel and not edge pixel;

classifying each pixel that is identified as said not edge pixel as one of a texture pixel and not texture pixel;

setting an edge block flag to a predetermined state upon said digitally noise reduced block of pixels having a predetermined number of pixels classified as edge pixels;

setting a texture block flag to a predetermined state upon said digitally noise reduced block of pixels having a predetermined number of pixels classified as texture pixels; and

filtering said digitally noise reduced block of pixels using a spatially-variant filter when any one of said block texture flag, said block edge flag, and said block replenish flag is set to said predetermined state.

39. A transform artifact reduction method for decoded video pixel data as in claim 38 wherein said identifying each pixel in said digitally noise reduced block of pixels as one of said edge pixel and said not edge pixel further comprises:

comparing a pixel gradient for a pixel with a threshold; and

identifying said pixel as said edge pixel upon said pixel gradient being greater than said threshold.

40. A transform artifact reduction method for decoded video pixel data as in claim 39 wherein said comparing said pixel gradient for said pixel further comprises:

comparing said pixel gradient with an adaptive edge threshold.

41. A transform artifact reduction method for decoded video pixel data as in claim 40 wherein said adaptive edge threshold is an adaptive texture threshold.

42. A transform artifact reduction method for decoded video pixel data as in claim 39 wherein said comparing said pixel gradient for said pixel further comprises:

comparing said pixel gradient with a luminance threshold.

43. A transform artifact reduction method for decoded video pixel data as in claim 39 wherein said comparing said pixel gradient for said pixel further comprises:

comparing said pixel gradient with both an adaptive edge threshold, and a luminance threshold.

44. A transform artifact reduction method for decoded video pixel data as in claim 43 wherein said identifying said pixel as said edge pixel upon said pixel gradient being greater than said threshold further comprises:

identifying said pixel as an edge pixel only upon said pixel gradient being greater than said adaptive edge threshold, and being greater than said luminance threshold.

45. A transform artifact reduction method for decoded video pixel data as in claim 43 further comprising prior to said comparing said pixel gradient:

generating said luminance threshold.

46. A transform artifact reduction method for decoded video pixel data as in claim 45 wherein said generating said luminance threshold further comprises:

generating a background luminance measure for said pixel wherein said luminance threshold is proportional to said background luminance measure.

47. A transform artifact reduction method for decoded video pixel data as in claim 46 wherein said generating said background luminance measure for said pixel further comprises:

averaging luminance components in a window about said pixel.

48. A transform artifact reduction method for decoded video pixel data as in claim 47 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

49. A transform artifact reduction method for decoded video pixel data as in claim 46 wherein said generating said luminance threshold further comprises:

adjusting said background luminance measure to obtain an adjusted background luminance measure.

50. A transform artifact reduction method for decoded video pixel data as in claim 49 wherein said generating said luminance threshold further comprises:

defining said luminance threshold as the maximum of a minimum luminance threshold and said adjusted background luminance measure.

51. transform artifact reduction method for decoded video pixel data as in claim 40 further comprising prior to said comparing said pixel gradient with said adaptive edge threshold:

generating said pixel gradient for said pixel.

52. A transform artifact reduction method for decoded video pixel data as in claim 51 wherein said generating said pixel gradient further comprises:

generating a plurality of spatial gradients for said pixel using pixels in a window about said pixel.

53. A transform artifact reduction method for decoded video pixel data as in claim 52 wherein said window is a three-by three pixel window with said pixel centered in said three-by-three pixel window.

54. A transform artifact reduction method for decoded video pixel data as in claim 52 wherein said plurality of spatial gradients comprise two spatial gradients.

55. A transform artifact reduction method for decoded video pixel data as in claim 52 wherein said plurality of spatial gradients comprise four spatial gradients.

56. A transform artifact reduction method for decoded video pixel data as in claim 52 wherein said generating said pixel gradient further comprises:

combining said plurality of spatial gradients to obtain said pixel gradient.

57. A transform artifact reduction method for decoded video pixel data as in claim 40 further comprising prior to said comparing said pixel gradient with said adaptive edge threshold:

generating said adaptive edge threshold.

58. A transform artifact reduction method for decoded video pixel data as in claim 57 wherein said generating said adaptive edge threshold further comprises:

generating an edge estimator for a plurality of edges through said pixel to obtain a plurality of edge estimators wherein pixels in a window about said pixel are used to generate said plurality of edge estimators.

59. A transform artifact reduction method for decoded video pixel data as in claim 58 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

60. A transform artifact reduction method for decoded video pixel data as in claim 58 wherein said plurality of edge estimators comprise two pixel texture estimators.

61. A transform artifact reduction method for decoded video pixel data as in claim 58 wherein said plurality of edge estimators comprise four pixel texture estimators.

62. A transform artifact reduction method for decoded video pixel data as in claim 58 wherein generating said adaptive edge threshold further comprises:

combining said plurality of edge estimators to obtain said adaptive edge threshold.

63. A transform artifact reduction method for decoded video pixel data as in claim 38 wherein said classifying each pixel that is identified as said not edge pixel as one of said texture pixel and said not texture pixel further comprises:

combining a plurality of edge estimators to obtain a texture estimator.

64. A transform artifact reduction method for decoded video pixel data as in claim 63 wherein said plurality of edge estimators were generated in said identifying each pixel as one of said edge pixel and said not edge pixel.

65. A transform artifact reduction method for decoded video pixel data as in claim 63 wherein said classifying each pixel that is identified as said not edge pixel as one of said texture pixel and said not texture pixel further comprises:

comparing said texture estimator to a texture pixel threshold wherein upon said texture estimator being greater than said texture pixel threshold, said pixel is classified as a texture pixel.

66. A transform artifact reduction method for decoded video pixel data as in claim 38 wherein said filtering of said digitally noise reduced block of pixels further comprises:

classifying each pixel as one of an edge pixel, an edge boundary pixel and a shade pixel.

67. A transform artifact reduction method for decoded video pixel data as in claim 66 wherein in said filtering of said digitally noise reduced block of pixels, a pixel is classified as said edge pixel if said pixel was identified as said edge pixel.

68. A transform artifact reduction method for decoded video pixel data as in claim 66 wherein said pixel classified as said edge pixel is unchanged by said filtering.

69. A transform artifact reduction method for decoded video pixel data as in claim 67 wherein said pixel is classified as said edge border pixel if (i) at least three pixels in a window of pixels about said pixel are identified as said not edge pixel;

and (ii) said pixel is in a line of pixels in said window and each pixel in the line is identified as said not edge pixel.

70. A transform artifact reduction method for decoded video pixel data as in claim 69 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

71. A transform artifact reduction method for decoded video pixel data as in claim 66 wherein in said filtering of said digitally noise reduced block of pixels, a pixel classified as an edge border pixel is filtered in a one-dimensional filter that processes said line of pixels.

72. A transform artifact reduction method for decoded video pixel data as in claim 71 where said one-dimensional filter is a low pass filter.

73. A transform artifact reduction method for decoded video pixel data as in claim 66 wherein said pixel is classified as said shade pixel if each pixel in a window of pixels about said pixel and said pixel are identified as said not edge pixels.

74. A transform artifact reduction method for decoded video pixel data as in claim 73 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

75. A transform artifact reduction method for decoded video pixel data as in claim 66 wherein said pixel classified as said shade pixel is filtered in a two-dimensional filter that processes said window of pixels.

76. A transform artifact reduction method for decoded video pixel data as in claim 75 where said two-dimensional filter is a low pass filter.

77. An edge detection method for transform artifact reduction comprising:

comparing a pixel gradient for a pixel with a luminance threshold;

setting an edge flag for said pixel in an edge map upon said pixel gradient being greater than said threshold;

repeating said comparing and setting processes for each pixel in a block of pixels.

78. An edge detection method for transform artifact reduction as in claim 77 further comprising:

comparing said pixel gradient with an adaptive edge threshold.

79. An edge detection method for transform artifact reduction as in claim 78 wherein said adaptive edge threshold is an adaptive texture threshold.

80. An edge detection method for transform artifact reduction as in claim 78 wherein said setting said edge flag for said pixel in said edge map upon said pixel gradient being greater than said threshold further comprises:

setting said edge flag only upon said pixel gradient being greater than said adaptive edge threshold, and being greater than said luminance threshold.

81. An edge detection method for transform artifact reduction as in claim 77 further comprising prior to said comparing said pixel gradient for said pixel with said luminance threshold:

generating said luminance threshold.

82. An edge detection method for transform artifact reduction as in claim 81 wherein said generating said luminance threshold further comprises:

generating a background luminance measure for said pixel wherein said luminance threshold is proportional to said background luminance measure.

83. An edge detection method for transform artifact reduction as in claim 82 wherein said generating a! said background luminance measure for said pixel further comprises:

averaging luminance components in a window about said pixel.

84. An edge detection method for transform artifact reduction as in claim 83 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

85. An edge detection method for transform artifact reduction as in claim 82 wherein said generating said luminance threshold further comprises:

adjusting said background luminance measure to obtain an adjusted background luminance measure.

86. An edge detection method for transform artifact reduction as in claim 85 wherein said generating said luminance threshold further comprises:

defining said luminance threshold as the maximum of a minimum luminance threshold and said adjusted background luminance measure.

87. An edge detection method for transform artifact reduction as in claim 78 further comprising prior to said comparing said pixel gradient with said adaptive edge threshold:

generating said pixel gradient for said pixel.

88. An edge detection method for transform artifact reduction as in claim 87 wherein said generating said pixel gradient further comprises:

generating a plurality of spatial gradients for said pixel using pixels in a window about said pixel.

89. An edge detection method for transform artifact reduction as in claim 88 wherein said window is a three-by three pixel window with said pixel centered in said three-by-three pixel window.

90. An edge detection method for transform artifact reduction as in claim 88 wherein said plurality of spatial gradients comprise two spatial gradients.

91. An edge detection method for transform artifact reduction as in claim 88 wherein said plurality of spatial gradients comprise four spatial gradients.

92. An edge detection method for transform artifact reduction as in claim 88 wherein said generating pixel gradient further comprises:

combining said plurality of spatial gradients to obtain said pixel gradient.

93. An edge detection method for transform artifact reduction as in claim 78 further comprising said pixel gradient with said adaptive edge threshold:

generating said adaptive edge threshold.

94. An edge detection method for transform artifact as in claim 93 wherein said generating said adaptive edge threshold further comprises:

generating an edge estimators for a plurality of edges through said pixel to obtain a plurality of edge estimators wherein pixels in a window about said pixel are used to generate said plurality of edge estimators.

95. An edge detection method for transform artifact reduction as in claim 94 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

96. An edge detection method for transform artifact reduction as in claim 94 wherein said plurality of edge estimators comprise two pixel texture estimators.

97. An edge detection method for transform artifact reduction as in claim 94 wherein said plurality of edge estimators comprise four pixel texture estimators.

98. An edge detection method for transformation artifact reduction as in claim 94 wherein said generating said adaptive edge threshold further comprises:

combining said plurality of edge estimators to obtain said adaptive edge threshold.

99. A spatially-variant filtering method to reduce transform artifacts comprising:

assigning a pixel in a block of pixels one classification in a plurality of edge classifications using edge flags in an edge map for pixels in a window about said pixel wherein an edge flag for a pixel is set in said edge map to indicate said pixel is in an edge of a video image; and

filtering each pixel in said block of pixels based upon said assigned classification to reduce transform artifacts in said video image.

100. A spatially-variant filtering method to reduce transform artifacts as in claim 99 wherein said plurality of edge classifications includes an edge classification, and said pixel is assigned the edge classification if said edge flag for said pixel in said edge map indicates that said pixel is an edge pixel.

101. A spatially-variant filtering method to reduce transform artifacts as in claim 100 wherein a pixel assigned said edge classification is unchanged by said filtering of said pixel.

102. A spatially-variant filtering method to reduce transform artifacts as in claim 99 wherein said plurality of edge classifications includes an edge border classification.

103. A spatially-variant filtering method to reduce transform artifacts as in claim 102 wherein said pixel is assigned said edge border classification if (i) the edge flag for at least three pixels in a window of pixels about said pixel are not set;

and (ii) said pixel is in a line of pixels in said window and the edge flag for each pixel in the line is not set.

104. A spatially-variant filtering method to reduce transform artifacts as in claim 103 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

105. A spatially-variant filtering method to reduce transform artifacts as in claim 103 wherein a pixel assigned said edge border classification is filtered in a one-dimensional filter that processes said line of pixels.

106. A spatially-variant filtering method to reduce transform artifacts as in claim 105 where said one-dimensional filter is a low pass filter.

107. A spatially-variant filtering method to reduce transform artifacts as in claim 99 wherein said plurality of edge classifications includes a shade classification.

108. A spatially-variant filtering method to reduce transform artifacts as in claim 107 wherein said pixel is assigned said shade classification if the edge flag for said pixel and each pixel in a window of pixels about said pixel is not set.

109. A spatially-variant filtering method to reduce transform artifacts as in claim 108 wherein said window is a three-by-three pixel window with said pixel centered in said three-by-three pixel window.

110. A spatially-variant filtering method to reduce transform artifacts as in claim 108 wherein a pixel assigned said shade classification is filtered in a two-dimensional filter that processes said window of pixels.

111. A spatially-variant filtering method to reduce transform artifacts as in claim 110 where said two-dimensional filter is a low pass filter.