Browse Prior Art Database

Simplified Motion Compensation Algorithms

IP.com Disclosure Number: IPCOM000109182D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-23
Document File: 2 page(s) / 89K

Publishing Venue

IBM

Related People

Galand, C: AUTHOR [+4]

Abstract

In a motion compensation-based video coder, each image is coded by prediction with reference to the previously reconstructed image. The predictor includes an activity detector to locate the active blocks, a motion estimator to predict a motion vector for each active block, and a motion compensation mechanism to reconstruct the image. The coordinates of the active blocks and associated motion vectors, as well as the output of the quantizer, are transmitted to the decoder to reconstruct the image sequences.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Simplified Motion Compensation Algorithms

       In a motion compensation-based video coder, each image is
coded by prediction with reference to the previously reconstructed
image.  The predictor includes an activity detector to locate the
active blocks, a motion estimator to predict a motion vector for each
active block, and a motion compensation mechanism to reconstruct the
image.  The coordinates of the active blocks and associated motion
vectors, as well as the output of the quantizer, are transmitted to
the decoder to reconstruct the image sequences.

      Motion compensation is a powerful method to reduce the
redundancy of video sequences, but usually leads to high complexity
coders, because for each block of pixels in the new image, the method
looks for the block in the previously reconstructed image that
provides the best match.  One way to reduce the complexity is to
split the image in sub-bands, to down-sample the sub-bands according
to their bandwidth, and to implement the motion compensation
algorithms on the sub-sampled lower sub-band.

      In the following, we consider video sequences of 256x256
pixels, at a rate of 25 images per second.  Let us denote I(n) the
image at time nt.  The image I(n) is sub-sampled by a factor of 2.
The decimated image is then filtered by a half-band low-pass filter,
and down-sampled by a factor of 2.  The resulting decimated image
D(n) only contains 64x64 pixels.  The motion compensation is
implemented in two step...