Image analysis – Image compression or coding – Interframe coding
Reexamination Certificate
1996-06-05
2002-08-06
Johnson, Timothy M. (Department: 2623)
Image analysis
Image compression or coding
Interframe coding
C375S240160
Reexamination Certificate
active
06430316
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to video compression methods and apparatus, in particular in the context of motion vector compensation techniques.
2. Description of the Prior Art
One digital compression standard, known as MPEG, was devised by and named after the Motion Picture Experts Group of the International Standards Organisation (ISO). In one form, known as MPEG1, early versions of the draft standard made use of forward prediction between past and current frames of a video signal. In later forms of MPEG1 and another known as MPEG2, the standard additionally makes use of bidirectional prediction, namely a combination of forward and backward prediction with the prediction made as an average of the forward and backward predictive frames.
Further details of the MPEG standard are set out in ISO/IEC 11172-2:1993(E) “Information Technology—Coding of Moving Pictures and Associated Audio for Digital Storage Media at up to about 1.5 Mbit/s”.
The use of motion compensation is an important aspect of MPEG coding. Motion compensation aims to minimize the error signal in difference frames by compensating for inter-frame movement. The process of motion compensation involves two separate processes:
i) Motion Vector Estimation (MVE); in which the motion offset between a frame pair is calculated; and
ii) Motion Vector Compensation (MVC); which uses the results of the MVE calculations to produce a difference signal with minimum artifacts.
There are several aspect of MVE and MVC which must be considered such as:
i) The type of motion permitted. In MPEG and other similar systems, the type of motion is limited simply to translational vectors in the ‘X’ and ‘Y’ axes. Researchers working in other fields have used other motion types, such as rotation, skew and size. The benefits of this work are not clear since using more complex shape matching may reduce the signal in the difference picture, but requires more data to transmit the shape parameters.
ii) The size of the macroblock is a compromise between smaller blocks having better coding efficiency but higher macroblock data rate, and larger blocks having worse coding efficiency but lower macroblock data rate (where a macroblock is the block size used for motion estimation and compensation).
iii) How to combine the results from Luminance (Y) and Chrominance (C) block matching. If completely separate vectors are used for Y and C motion vectors, the macroblock data rate will be doubled and there may be the possibility of picture disturbances created by poor matching of Y and C vectors. Combining the Y and C vectors will result in a lower macroblock data rate but the coding efficiency may drop slightly.
iv) The use of sub-pixel estimation and compensation. Often, motion between frames does not appear at integer pixel distances so some form of sub-pixel MVE and MVC will be beneficial. More vector data is generated to be transmitted/recorded and the data increase is very small but simple half-pixel estimation requires at least four times the computational power of a pixel based estimator and this must be considered.
There is a need for further developments in motion compensation coding with emphasis on low bit-rate systems in the range 64 Kbps to 2 Mbps.
SUMMARY OF THE INVENTION
It is an object of the invention to provide a video compression method involving motion vector compensation in which the problem of macroblock size compromise is alleviated.
It is another object of the invention to provide a video compression method involving motion vector compensation which is applicable to low bit-rate systems.
According to the invention there is provided a video compression method involving motion vector compensation, the method comprising:
defining a macroblock structure over a video picture to which motion vector analysis may be applied; and
applying a soft window to the macroblock structure such that adjacent windows overlap, with the edges of each window forming progressively increasing pixel weightings from the edge inwards, the soft window being larger than each macroblock in the structure.
The pixel weightings are preferably selected so that each overlapped pixel arising from overlapping windows has a summed weighting equal to a constant.
The soft window is preferably 50% larger in each dimension than a macroblock of the structure, and the tapering part of each window edge preferably comprises one third of the total length. Where the macroblock sizes differ for chrominance and luminance, the window sizes may differ proportionately.
In a preferred embodiment of the invention, the window with soft edges provides motion compensation with much better results on many types of test material. Although errors at higher video data rates may not be so visible, applying the window still leads to a general reduction in picture distortions. The window is simple to apply and the window coefficients (or weightings) are easy to use.
The above, and other objects, features and advantages of this invention will be apparent from the following detailed description of illustrative embodiments which is to be read in connection with the accompanying drawings.
REFERENCES:
patent: 4862261 (1989-08-01), Tanaka
patent: 5206723 (1993-04-01), Parke
patent: 5400087 (1995-03-01), Uramoto et al.
patent: 5453801 (1995-09-01), Kim
patent: 5512962 (1996-04-01), Homma
patent: 5523850 (1996-06-01), Kanda et al.
patent: 5526053 (1996-06-01), Dorricott et al.
patent: 5535288 (1996-07-01), Chen et al.
patent: 5579050 (1996-11-01), Jung
patent: 5602593 (1997-02-01), Katto
patent: 6005627 (1999-12-01), Odaka et al.
Watanabe et al., Windowed Motion Compensation, 1991, p 582-589, SPIE V 1605 Visual Communications and Image Processing '91: Visual Communication.*
Sullivan et al., Methods of reduced-complexity overlapped block motion compensation, Nov. 1994, pp. 957-961, IEEE ICIP-94, vol. 2.*
Katto et al., A wavelet codec with overlapped motion compensation for very low bit-rate environment, Jun. 1994, pp 328-338, IEEE Transactions on Circuits and Systems for Video Technology.*
Orchard et al., Overlapped block motion compensation: an estimation-theoretic approach, Sep. 1994, IEEE Transactions on Image Processing, vol. 3, Iss: 5, p. 693-9.*
Nogaki et al., An overlapped block motion compensation for high quality motion picture coding, May 1992, IEEE International Symposium on Circuits and Systems p. 184-7 vol. 1.*
Katto et al., An analytical framework for overlapped motion compensation, May 1995, p. 2189-2192, 1995 International Conference on Acoustics, Speech, and Signal Processing.
Fromer William S.
Fromer Lawrence & Haug LLP
Johnson Timothy M.
Shallenburger Joe H.
LandOfFree
Motion vector compensation using overlapping weighted windows does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Motion vector compensation using overlapping weighted windows, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion vector compensation using overlapping weighted windows will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2948372