Appeal No. 2006-1605
Application No. 09/470,741
in the least squares sense as they minimize the mean square
error between a reference block and a block obtained through
low-resolution MC. The results which have been derived in
[8] assume that a spatial domain filter, x, is applied to
incoming macroblocks to achieve the down-conversion. The
scheme shown in Fig. 3(a) illustrates the process by which
reference blocks are obtained. First, full-resolution
motion compensation is performed on macroblocks, a, b, c,
and d to yield, h. To execute this process, the filters
Sa(r) , Sb(r), Sc(r), and Sd(r) are used. Basically, these
filters represent the masking/averaging operations of the
motion compensation in a matrix form. More on the
composition of these filters can be found in the appendix.
Once h is obtained, it is down-converted to h via the
spatial filter, x:
(1) h = xh.
2. Vetro states, at page 11, Section 4.3, first paragraph, the
following:
To perform low-resolution MC, the high-definition
motion vector must specify the neighborhood of blocks
in which to filter. Once these blocks are retrieved
from the frame store, a filtering operation is applied.
The set of filters which are used to operate on the
reference blocks are specified by the amount of overlap
as indicated by y1 and y2 (see appendix and Fig 12) and
the prediction mode. Once this filtering is performed,
the low-resolution prediction can be added to the
residual component to form the reconstructed block.
3. Ng states, in column 5, line 60- column 6, line 7, the
following:
FIG. 5 illustrates a further embodiment which produces
improved images over the FIG. 4 embodiment. The
improvement results because advantage is taken of the
total motion vectors not truncated motion vectors or
the affects of truncating memory addresses to the VRAM
315. In FIG. 5 an interpolator 319 is interposed
11
Page: Previous 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Next
Last modified: November 3, 2007