Databases Reference
In-Depth Information
Example19.3.2:
Let us again try to predict the second frame of Example 10.3.1 using motion compensation.
We divide the image into blocks and then predict the second frame from the first in the manner
described above. Figure
19.3
shows the blocks in the previous frame that were used to predict
some of the blocks in the current frame.
F I GU R E 19 . 3
Motion-compensated prediction.
Notice that in this case all that needs to be transmitted to the receiver are the motion vectors.
The current frame is completely predicted by the previous frame.
We have been describing motion compensation where the displacement between the block
being encoded and the best matching block is an integer number of pixels in the horizontal
and vertical directions. There are algorithms in which the displacement is measured in half
pixels. In order to do this, pixels of the coded frame being searched are interpolated to obtain
twice as many pixels as in the original frame. This “doubled” image is then searched for the
best matching block.
The doubled image is obtained as follows: Consider Table
19.1
. In this image
A
,
B
,
C
,
and
D
are the pixels of the original frame. The pixels
h
1
,
h
2
,v
1
, and
v
2
are obtained by
interpolating between the two neighboring pixels:
T A B L E 19 . 1
“Doubled” image.
A
h
1
B
v
1
c
v
2
C
h
2
D
