Efficient motion vector coding for video compression

• US 20050036759A1
• Filed: 06/30/2004
• Published: 02/17/2005
• Est. Priority Date: 11/30/1998
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Lee, Ming-Chieh, Lin, Chih-Lung (Bruce)
• Granted Patent: US 7,408,990 B2
• Time in Patent Office: Days
• US Class Current: 386/331
• CPC Class Codes: H04N 19/51 Motion estimation or motion compensation
• CPC Class Codes: H04N 19/52 by predictive encoding

Abstract

Video coding efficiency is improved by jointly coding the x and y components of motion vectors with a single variable length code. The motion vector components for a block of pixels are predicted based on motion vectors of neighboring blocks of pixels. The predicted x and y components are then jointly coded by assigning a single variable length code corresponding to the pair of components, rather than a separate code for each component. If the x and y components do not have a corresponding entry in the coding table, they are coded with an escape code followed by fixed length codes.

Claims

1-19. -19. (Canceled)

20. A method comprising:

selecting a variable length code table from among plural available variable length code tables, wherein each of the plural available variable length code tables has a different set of plural variable length codes used for encoding and decoding motion vector differential information;

computing a motion vector predictor for a motion vector, wherein the motion vector predictor includes a horizontal motion vector predictor component and a vertical motion vector predictor component, and wherein the motion vector includes a horizontal motion vector component and a vertical motion vector component;

computing a horizontal motion vector differential component based at least in part upon difference between the horizontal motion vector component and the horizontal motion vector predictor component;

computing a vertical motion vector differential component based at least in part upon difference between the vertical motion vector component and the vertical motion vector predictor component; and

using the selected variable length code table, producing a first variable length code from the set of plural variable length codes of the selected variable length code table, wherein the first variable length code jointly represents information used for decoding the horizontal motion vector differential component and for decoding the vertical motion vector differential component.

21. The method of claim 20 wherein the motion vector is for a current block or macroblock of a video picture, and wherein the computing the motion vector predictor depends at least in part on location of the current block or macroblock in the video picture.

22. The method of claim 20 wherein the horizontal motion vector predictor component is the median of three neighbor horizontal motion vector components, and wherein the vertical motion vector predictor component is the median of three neighbor vertical motion vector components.

23. The method of claim 20 wherein a first neighbor to the left of a current macroblock, a second neighbor above the current macroblock, and a third neighbor above and diagonally adjacent to the current macroblock with a position to the left or right depending on spatial location of the current macroblock, each have one of three neighbor horizontal motion vector components and one of three neighbor vertical motion vector components.

24. The method of claim 20 further comprising:

producing first additional information used for decoding the horizontal motion vector differential component; and

producing second additional information used for decoding the vertical motion vector differential component.

25. The method of claim 24 wherein the first variable length code indicates the presence of the first and second additional information, and wherein the first variable length code, the first additional information, and the second additional information collectively have variable length from motion vector to motion vector.

26. The method of claim 20 wherein the first variable length code indicates the absence of first additional information used for decoding the horizontal motion vector differential component and second additional information used for decoding the vertical motion vector differential component.

27. The method of claim 20 wherein the motion vector is for a block.

28. The method of claim 20 wherein the motion vector is for a macroblock.

29. The method of claim 20 wherein the selected variable length code table is adapted for a particular target bit rate and/or a particular kind of content.

30. A method comprising:

for each macroblock of plural macroblocks of a video picture,

receiving a variable length code that jointly represents information used for reconstructing a horizontal motion vector differential component for a motion vector of the macroblock and a vertical motion vector differential component for the motion vector of the macroblock; and

reconstructing the horizontal motion vector differential component and the vertical motion vector differential component based at least in part on the variable length code, including looking up the variable length code in one of plural variable length code tables for motion vector differential information, wherein each of the plural variable length code tables has a different set of plural variable length codes.

31. The method of claim 30 further comprising:

for each macroblock of the plural macroblocks of the video picture,

computing a motion vector predictor for the motion vector for the macroblock, wherein the motion vector predictor includes a horizontal motion vector predictor component and a vertical motion vector predictor component, and wherein the motion vector includes a horizontal motion vector component and a vertical motion vector component; and

reconstructing the motion vector for the macroblock, including combining the horizontal motion vector differential component with the horizontal motion vector predictor component and combining the vertical motion vector differential component with the vertical motion vector predictor component.

32. The method of claim 31 wherein the computing the motion vector predictor depends at least in part on location of the macroblock in the video picture.

33. The method of claim 31 wherein the horizontal motion vector predictor component is the median of three neighbor horizontal motion vector components, and wherein the vertical motion vector predictor component is the median of three neighbor vertical motion vector components.

34. The method of claim 31 wherein a first neighbor to the left of a current macroblock, a second neighbor above the current macroblock, and a third neighbor above and diagonally adjacent to the current macroblock with a position to the left or right depending on spatial location of the current macroblock, each have one of three neighbor horizontal motion vector components and one of three neighbor vertical motion vector components.

35. The method of claim 30 wherein the video picture is a video frame.

36. The method of claim 30 further comprising receiving first and second additional information after the variable length code, wherein the reconstructing the horizontal motion vector differential component is further based at least in part on the first additional information, wherein the reconstructing the vertical motion vector differential component is further based at least in part on the second additional information, and wherein the variable length code the first and second additional information collectively have variable length from motion vector to motion vector.

37. The method of claim 36 wherein the variable length code indicates the presence of the first and second additional information.

38. The method of claim 30 wherein the variable length code indicates the absence of first additional information for the horizontal motion vector differential component and second additional information for the vertical motion vector differential component.

39. The method of claim 30 wherein the selected variable length code table is adapted for a particular target bit rate and/or a particular kind of content.

40. A computer-readable medium storing computer-executable instructions for causing a computer system programmed thereby to perform the method of claim 30.

41. A method comprising:

for each block of plural blocks of a video picture,

receiving a variable length code that jointly represents information used for reconstructing a horizontal motion vector differential component for a motion vector for the block and a vertical motion vector differential component for the motion vector for the block; and

reconstructing the horizontal motion vector differential component and the vertical motion vector differential component based at least in part on the variable length code, including looking up the variable length code in one of plural variable length code tables for motion vector differential information, wherein each of the plural variable length code tables has a different set of plural variable length codes.

42. The method of claim 41 wherein the video picture is a video frame.

43. The method of claim 41 further comprising receiving first and second additional information after the variable length code, wherein the reconstructing the horizontal motion vector differential component is further based at least in part on the first additional information, and wherein the reconstructing the vertical motion vector differential component is further based at least in part on the second additional information.

44. The method of claim 43 wherein the variable length code indicates the presence of the first and second additional information, and wherein the variable length code the first and second additional information collectively have variable length from motion vector to motion vector.

45. The method of claim 41 wherein the variable length code indicates the absence of first additional information for the horizontal motion vector differential component and second additional information for the vertical motion vector differential component.

46. The method of claim 41 wherein the selected variable length code table is adapted for a particular target bit rate and/or a particular kind of content.

47. A computer-readable medium storing computer-executable instructions for causing a computer system programmed thereby to perform the method of claim 41.

48. A method comprising:

receiving a variable length code that jointly represents information used for reconstructing a horizontal motion vector differential component for a motion vector and for reconstructing a vertical motion vector differential component for the motion vector;

decoding the variable length code with one of plural alternative variable length code tables for motion vector differential information, wherein each of the plural alternative variable length code tables uses a different set of plural variable length codes;

receiving first additional information;

reconstructing the horizontal motion vector differential component based at least in part on the first additional information;

receiving second additional information; and

reconstructing the vertical motion vector differential component based at least in part on the second additional information.

49. The method of claim 48 further comprising

computing a motion vector predictor for the motion vector, wherein the motion vector predictor includes a horizontal motion vector predictor component and a vertical motion vector predictor component, and wherein the motion vector includes a horizontal motion vector component and a vertical motion vector component;

and

reconstructing the motion vector, including combining the horizontal motion vector differential component with the horizontal motion vector predictor component and combining the vertical motion vector differential component with the vertical motion vector predictor component.

50. The method of claim 49 wherein the motion vector is for a current block or macroblock of a video picture, and wherein the computing the motion vector predictor depends at least in part on location of the current block or macroblock in the video picture.

51. The method of claim 49 wherein the horizontal motion vector predictor component is the median of plural neighbor horizontal motion vector components, and wherein the vertical motion vector predictor component is the median of plural neighbor vertical motion vector components.

52. The method of claim 48 wherein the variable length code indicates the presence of the first and second additional information.

53. The method of claim 52 wherein the variable length code the first and second additional information collectively have variable length from motion vector to motion vector.

54. The method of claim 48 wherein the motion vector is for a block.

55. The method of claim 48 wherein the motion vector is for a macroblock.

56. The method of claim 48 wherein the selected variable length code table is adapted for a particular target bit rate and/or a particular kind of content.

57. A computer-readable medium storing computer-executable instructions for causing a computer system programmed thereby to perform the method of claim 48.