Video encoder/decoder system

• US 5,805,228 A
• Filed: 08/09/1996
• Issued: 09/08/1998
• Est. Priority Date: 08/09/1996
• Status: Expired due to Term

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: US Robotics Access Corp. (US Robotics Incorporated)
• Inventors: Smith, David L., Truong, Kwan K., Richardson, Craig H., Hodges, Chris J. M., Proctor, John J.
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US08/689,444
• Time in Patent Office: 760 Days
• Field of Search: 348/417, 348/418, 348/699, 348/411, 348/412, 348/408, 348/414, 348/416, 348/422, 348/700, 382/240, 382/253
• US Class Current: 375/240.22
• CPC Class Codes: G06T 9/008 Vector quantisation
• CPC Class Codes: H04N 19/94 Vector quantisation

Abstract

A method and an apparatus for encoding an image signal. The apparatus includes an acquisition module disposed to receive the image signal. A first processor is coupled to the acquisition module. At least one encoder processor is coupled to the first processor. The at least one encoder processor produces an encoded image signal under control of the first processor. The method includes the steps of converting an input image signal into a predetermined digital format and transferring the digital format image signal to at least one encoder processor. The method further includes the step of applying, at the at least one encoder processor, a hierarchical vector quantization compression algorithm to the digitized image signal. At the next step, a resultant encoded bit stream generated by the application of the algorithm is collected. The method and apparatus of the present invention may be used in conjunction with an ordinary modem to transmit and/or receive audio, video sequences or still images.

Claims

1. An apparatus for encoding an image signal, comprising:

an acquisition module disposed to receive said image signal;

a first processor coupled to said acquisition module; and

two encoder processors coupled in parallel to said first processor, wherein said two encoder processors produce an encoded image signal under control of said first processor.

2. An apparatus as claimed in claim 1, wherein said acquisition module comprises means for converting an analog signal into a digital signal.

3. An apparatus as claimed in claim 2, wherein said digital signal is in a digital YUV format.

4. An apparatus as claimed in claim 3, wherein a U component and a V component of said YUV formatted digital signal are subsampled by a factor of four relative to a Y component of said YUV formatted digital signal.

5. An apparatus as claimed in claim 3, further comprising means for converting said YUV formatted digital signal into an RGB formatted signal, said converting means being coupled to said first processor.

6. An apparatus as claimed in claim 5, further comprising a display device coupled to said converting means.

7. An apparatus as claimed in claim 1, wherein said two encoder processors apply a hierarchical vector quantization compression algorithm to said signal.

8. An apparatus as claimed in claim 7, wherein said two encoder processors comprise digital signal processors.

9. An apparatus as claimed in claim 1, further comprising a data storage device coupled between said first processor and said two encoder processors.

10. An apparatus as claimed in claim 9, wherein said data storage device comprises a direct memory access interface.

11. An apparatus as claimed in claim 1, wherein said first processor performs at least one of video decoding, data acquisition and control.

12. An apparatus as claimed in claim 1, further comprising means for spatially partitioning said image signal among said two encoder processors.

13. An apparatus as claimed in claim 12, wherein said partitioning means is responsive to said encoded image signal.

14. An apparatus as claimed in claim 12, wherein said two encoder processors receive unequal partitions of said image signal.

15. An apparatus as claimed in claim 12, wherein said partitioning means comprises means for calculating a mean squared error or mean absolute error of a portion of said image signal.

16. An apparatus as claimed in claim 4, wherein said two encoder processors comprise means for detecting a scene change.

17. An apparatus as claimed in claim 16, wherein said detecting means calculates a difference between an incoming image signal and a previously coded image signal.

18. An apparatus as claimed in claim 17, wherein said detecting means compares said calculated difference to a threshold.

19. An apparatus as claimed in claim 18, wherein said detecting means calculates a mean squared error between said Y component of said incoming image signal and said Y component of said previously coded image signal.

20. An apparatus as claimed in claim 18, wherein said two encoder processors encode said incoming image signal as a still image when said difference exceeds said threshold.

21. An apparatus as claimed in claims 7 or 20, further comprising a memory structure containing a vector quantization codebook, said memory structure being coupled to said two encoders.

22. An apparatus as claimed in claim 1, further comprising a buffer coupled between said two encoder processors and said first processor.

23. An apparatus as claimed in claim 22, wherein said encoded image signal produced by said two encoder processors accumulates in said buffer for transmission to said first processor.

24. An apparatus as claimed in claim 22, further comprising a mean squared error accelerator coupled to said two encoder processors.

25. An apparatus as claimed in claim 24, wherein said mean squared error accelerator comprises a field programmable gate array or an application specific integrated circuit.

26. An apparatus as claimed in claim 22, further comprising a mean absolute error accelerator coupled to said two encoder processors.

61. An apparatus as in claim 26, wherein said mean absolute error accelerator comprises a field programmable gate array or application specific integrated circuit.

27. A method for generating a compressed video signal, comprising the steps of:

converting an input image signal into a predetermined digital format;

transferring said digital format image signal to a parallel processing encoder, said parallel processing encoder having separate processing elements that are arranged in parallel to receive different portions of said digitized image signal;

applying, at said encoder, a hierarchical vector quantization compression algorithm to the different portions of said digitized image signal; and

collecting a resultant encoded bit stream generated by said application of said algorithm.

28. A method as claimed in claim 27, further comprising the step of converting said resultant encoded bit stream into a digital signal having a RGB format.

29. A method as claimed in claim 28, wherein said predetermined digital format comprises data for a plurality of macro rows.

30. A method as claimed in claim 29, wherein said step of applying, at said encoder, said hierarchical vector quantization compression algorithm to said digitized image signal comprises sequentially applying said algorithm to each macro row of said plurality of macro rows.

31. A method as claimed in claim 27, wherein said digital image signal is transferred to a memory that is accessible to more than one encoder processor.

32. A method as claimed in claim 31, wherein said step of applying said hierarchical vector quantization compression algorithm further comprises said steps of:

applying, at a first encoder processor, said algorithm to a first portion of said digital image signal; and

applying, at a second encoder processor, said algorithm to a second portion of said digital image signal.

33. A method as claimed in claim 32, wherein said first portion of said digital image signal is distinct from said second portion of said digital image signal.

34. A method as claimed in claim 33, further comprising the step of varying a size of said first portion and a size of said second portion in response to said resultant encoded bit stream or activity in the respective image parts.

35. A method as claimed in claim 34, further comprising the step of transferring a portion of said digital image signal between said first encoder processor and said second encoder processor.

36. A method as claimed in claim 35, further comprising the step of transferring mean squared error data corresponding to said portion of said digital image signal between said first encoder processor and said second encoder processor.

37. A method as claimed in claim 34, wherein said sizes are varied in accordance with a length of said resultant encoded bit stream or activity in the respective image parts.

38. A method as claimed in claim 34, wherein said sizes are varied in accordance with a flag bit within said resultant encoded bit stream.

39. A method as claimed in claim 27, further comprising the steps of:

storing said digital format image signal corresponding to said converted input image signal;

converting a second input image signal into said predetermined digital format;

calculating a difference between said stored input image signal and said second input image signal; and

comparing said difference to a threshold.

40. A method as claimed in claim 39, further comprising the step of encoding said second input image signal as a still image when said difference exceeds said threshold.

41. A method as claimed in claim 39, wherein said stored input image signal comprises a first plurality of image blocks, and said second input image signal comprises a second plurality of image blocks.

42. A method as claimed in claim 41, wherein said step of calculating said difference comprises calculating a mean squared error between an image block from said first plurality of image blocks and a corresponding image block from said second plurality of image blocks.

43. A method as claimed in claim 42, further comprising the step of estimating how many bits are used to encode each of said second plurality of image blocks based upon said calculated mean squared error.

44. A method as claimed in claim 43, further comprising the step of adjusting a threshold applied by said at least one encoder processor in accordance with said estimate.

45. A method as claimed in claim 27, wherein said resultant encoded bit stream is collected in a buffer coupled to said parallel processing encoder.

46. A method as claimed in claim 45, further comprising the step of transferring said resultant encoded bit stream from said buffer to a control processor when a predetermined portion of said input image signal is encoded.

47. A method as claimed in claim 46, further comprising the step of providing feedback to said parallel processing encoder, said feedback indicating fullness of said buffer.

48. A method as claimed in claim 47, further comprising the step of adjusting a threshold applied by said parallel processing encoder in accordance with said feedback.

49. A method as claimed in claim 27, wherein said step of applying, at said encoder, said hierarchical vector quantization compression algorithm to said digitized image signal comprises dividing said digitized image signal into a plurality of image blocks.

50. A method as claimed in claim 49, further comprising the steps of:

calculating a mean value for at least one of said plurality of image blocks; and

storing each calculated mean value in a scalar cache.

51. A method as claimed in claim 50, wherein said steps of calculating a mean value and storing said calculated mean value in said scalar cache are repeated at each level of said hierarchical vector quantization compression algorithm.

52. A method as claimed in claim 50, further comprising the step of encoding a mean value of an image block with a mean value stored in said scalar cache.

53. A method as claimed in claim 27, further comprising the steps of calculating an overall scene change mean squared error and comparing said calculated value to a first threshold before said step of applying said hierarchical vector quantization compression algorithm to said digitized image signal.

54. A method as claimed in claim 53, further comprising the step of predicting a motion vector based upon said digitized image signal if said calculated value does not exceed said first threshold.

55. A method as claimed in claim 54, further comprising the step of encoding an image block by its mean value if said calculated value does not exceed a second threshold.

56. A method as claimed in claim 55, further comprising the step of encoding an image block by a mean-removed vector quantizer if said calculated value does not exceed a third threshold.

57. A method as claimed in claim 27, wherein said step of applying, at said encoder, a hierarchical vector quantization compression algorithm to said digitized image signal comprises the steps of:

providing a motion vector cache and a vector quantization cache;

storing initially a set of most probable motion vectors in said motion vector cache and a set of most probable code vectors in said vector quantization cache;

encoding said digitized image signal in accordance with said hierarchical vector quantization compression algorithm; and

reinitializing said motion vector cache with said set of most probable motion vectors and reinitializing said vector quantization cache with said set of most probable code vectors.

59. A method as claimed in claim 57, further comprising the steps of:

providing a decoder having a decoder motion vector cache and a decoder vector quantization cache;

storing initially said set of most probable motion vectors in said decoder motion vector cache and said set of most probable code vectors in said decoder vector quantization cache;

decoding said resultant encoded bit stream; and

reinitializing said decoder motion vector cache with said set of most probable motion vectors and reinitializing said decoder vector quantization cache with said set of most probable code vectors.

60. A method as claimed in claim 59, wherein said decoder motion vector cache and said decoder vector quantization cache are reinitialized periodically during decoding.

62. A method as claimed in claim 27, wherein said step of applying, at said encoder, a hierarchical vector quantization compression to said digitized image signal comprises the steps of:

applying said hierarchical vector quantization compression algorithm to a plurality of image blocks corresponding to said digitized image signal, wherein an image block from said plurality of image blocks is selected to be encoded in accordance with a scanning pattern; and

varying said scanning pattern when said algorithm is applied to a subsequent digitized image signal.

58. A method as claimed in 57, wherein said motion vector cache and said vector quantization cache are periodically reinitialized.

63. A method for compressing an image signal representing a series of image frames, comprising the steps of:

a) analyzing a first image frame by;

i) computing a mean value for each of a plurality of image blocks within said first image frame;

ii) storing said computed mean values in a scalar cache;

iii) providing a mean value quantizer comprising a predetermined number of quantization levels arranged between a minimum mean value and a maximum mean value stored in said scalar cache, said mean value quantizer producing a quantized mean value; and

iv) identifying each image block from said plurality of image blocks that is a low activity image block;

b) encoding each of said low activity image blocks with its corresponding quantized mean value; and

c) repeating steps a) and b) for a second frame of said image signal.

64. A method as claimed in claim 63, further comprising, after step b), said step of encoding remaining image blocks using a hierarchical vector quantization compression algorithm.

65. A method of motion estimation between a current image and a reference image in a video compression system, comprising the steps, in combination, of:

generating a first reference macro row, a second reference macro row, and a third reference macro row, said first, second and third reference macro rows corresponding to interpolated image data of said reference image;

storing said first reference macro row in a first memory location, said second reference macro row in a second memory location, and said third reference macro row in a third memory location;

generating a first motion estimation value corresponding to a first current and a second current image macro row of said current image;

copying said second reference macro row to said first memory location;

copying said third reference macro row to said second memory location;

generating a fourth reference macro row corresponding to said reference image;

copying said fourth reference macro row to said third memory location; and

generating a second motion estimation value corresponding to a third current image macro row of said current image.

66. A method as in claim 65, wherein said steps are repeated in combination.

67. A method as in claim 65, wherein said first, second, third and fourth reference macro rows are interpolated by a factor of two in the horizontal direction.

68. A method as in claim 66, wherein said first, second, third and fourth reference macro rows are interpolated by a factor of two in the horizontal direction.

69. A method as in claim 65, wherein said first, second, third and fourth reference macro rows are interpolated by a factor of two in the vertical direction.

70. A method as in claim 66, wherein said first, second, third and fourth reference macro rows are interpolated by a factor of two in the vertical direction.

77. A method as in claim 69, wherein said reference image comprises 96 horizontal rows of pixels.

71. A method of motion estimation between a current image and a reference image in a video compression system, comprising the steps, in combination, of:

generating a first reference macro row, a second reference macro row, and a third reference macro row, said first, second and third reference macro rows corresponding to interpolated image data of said reference image;

storing said first reference macro row in a first memory location, said second reference macro row in a second memory location, and said third reference macro row in a third memory location;

generating a first motion estimation value corresponding to a first current and a second current image macro row of said current image;

setting a pointer to said third memory location;

copying a fourth reference macro row to said first memory location;

generating a motion estimation value corresponding to a third current image macro row of said current image;

setting said pointer to said first memory location;

copying a fifth reference macro row to said second memory location;

generating a second motion estimation value corresponding to a fourth current image macro row of said current image;

setting said pointer to said second memory location;

copying a sixth reference macro row to said third memory location;

generating a third motion estimation value corresponding to a fifth current image macro row of said current image.

72. A method as in claim 71, wherein said steps are repeated in combination.

73. A method as in claim 71, wherein said first, second, third, fourth, fifth and sixth reference macro rows are interpolated by a factor of two in the horizontal direction.

74. A method as in claim 72, wherein said first, second, third, fourth, fifth and sixth reference macro rows are interpolated by a factor of two in the horizontal direction.

75. A method as in claim 71, wherein said first, second, third, fourth, fifth and sixth reference macro rows are interpolated by a factor of two in the vertical direction.

76. A method as in claim 72, wherein said first, second, third, fourth, fifth and sixth reference macro rows are interpolated by a factor of two in the vertical direction.

78. A method as in claim 75, wherein said first second memory location begins at the first row of said 96 horizontal rows, said second memory location begins at the 32nd row of said 96 rows, and said third memory location begins at the 64th row of said 96 rows.