Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1998-11-05
2004-08-17
Lee, Richard (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C375S240290
Reexamination Certificate
active
06778604
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to video compression, and more particularly to a motion compensation performance improvement in motion compensated video coding by removing redundant edge information.
Motion compensation is well known for digital video compression. Its main use is to reduce the temporal redundancy of a video sequence. In practice the motion vector precision is often limited by compression standards to, for example, one-half pixel precision. Standard motion compensation techniques search for a minimum residue within the confinement of the limited fractional pixel precision. A shifted reference block is subtracted from a current block of video data to form a motion compensated residue. The block residue is quantized and entropy coded along with a motion vector that defines the amount of the shift for the reference block. A decoder performs inverse entropy coding to recover the quantized residue which in turn is inverse quantized and added to the shifted reference block to form the current block.
Bernd Girod, in his article entitled “Motion-Compensating Prediction with Fractional-pel Accuracy”, IEEE Transactions on Communications, Vol. 41, No. 4 April 1993, studied the effect of fractional pixel accuracy on the efficiency of motion compensation prediction in conjunction with various spatial interpolation/prediction filters. Without exception the prediction error variance decreased monotonically with increasing motion compensation accuracy. He concluded that quarter pixel accuracy appears sufficient for broadcast television signals. However even with quarter pixel accuracy, the prediction error still contains a considerable amount of edge information. A compression system has to balance the cost of transmitting motion vectors against the amount of prediction error by choosing the appropriate accuracy for the motion vectors.
When the fractional pixel movement of the video data falls exactly onto a quantized motion vector grid, i.e., integer and/or half pixel movement, then the residue of the motion compensation is optimum both in the sense of minimizing residue values and visual distortion. However if the fractional pixel movement does not fall exactly onto the quantized motion vector grid, then the residue of the motion compensation contains redundant edge information. For example, shift an image by a small fraction, like a quarter pixel, and subtract the shifted image from the original image to form a difference image. Upon visual inspection the difference image contains mostly edge information. The edge information is similar to the original image and is redundant in the sense that human visual systems cannot easily tell the difference between the shifted and original images. Due to the fact that the residue of the motion compensation is often quantized heavily in order to achieve lower transmission rates, the redundant edge information is distorted. The resulting artifacts created are very visible if the situation persists for some short period of time. Such an often observed artifact in MPEG-2 encoded video is the shimmering of fine details when panning a scene.
What is desired is an improved motion compensation performance technique that minimizes redundant edge information even with the pixel precision limitation placed on the motion vector accuracy by a compression standard.
BRIEF SUMMARY OF THE INVENTION
Accordingly the present invention provides motion compensation performance improvement by minimizing redundant edge information. A current image from an input video signal, which may be either the entire picture or a sub-region of the picture, is compared with a prior reference image from the input video signal, which may correspondingly be either the entire picture or a sub-region of the picture, to determine a spatial shift value between the two images. The spatial shift value has an integer portion and a high precision fractional portion. The high precision fractional portion together with specified constants according to the particular compression standard is used to calculate a fractional shift value that aligns the current image with a quantized motion vector grid for the compression encoder. The current image is resampled using the fractional shift value before being input to the compression encoder, thus minimizing the redundant edge information.
The objects, advantages and other novel features of the present invention are apparent from the following detailed description when read in conjunction with the appended claims and attached drawing.
REFERENCES:
patent: 4757383 (1988-07-01), Tanaka
patent: 5109451 (1992-04-01), Aono et al.
patent: 5299019 (1994-03-01), Pack et al.
patent: 5317397 (1994-05-01), Odaka et al.
patent: 5361105 (1994-11-01), Iu
patent: 5542008 (1996-07-01), Sugahara et al.
patent: 5748247 (1998-05-01), Hu
patent: 6064450 (2000-05-01), Canfield et al.
Girod, “Motion-Compensating Prediction with Fractional-Pel Accuracy”, IEEE Transactions on Communications, vol. 41, No. 4, pp. 604-612, Apr. 1993.
Gray Francis I.
Lee Richard
Tektronix Inc.
LandOfFree
Motion compensation performance improvement by removing... 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 compensation performance improvement by removing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion compensation performance improvement by removing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3348884