Appeal No. 2006-1605
Application No. 09/470,741
between the VRAM 315 and the predictor 304. In
addition a two dimensional decimator 313, similar to
decimator 311, is interposed between the predictor 304
and the adder 312. The interpolator 319 accepts blocks
of data from the VRAM 315 and generates 8 by 8 blocks
which are coupled to the predictor. The predictor
couples the 8 by 8 blocks of data to the decimator 313
which subsamples the data back down to 4 by 4 blocks of
data in conformance with the data format applied to the
adder from decimator 311.
4. Bose states, in column 17, lines 4-15, the following:
An example of how motion compensation is carried out
with the memory map in Fig. 12 where the luminance or
4:2:2 chrominance data of a full frame macroblock Sp of
a P-picture, say in buffer 73B, is to be rewritten with
a 16x16 pel square 90 (superimposed on Fig. 12 for
illustration) of data from a reference I- or P-picture
that is displaced, as specified by a motion
compensation vector 88, at some relative vertical and
horizontal distance in buffer 73A. The motion vector
88 may specify, for example, that the 16x16 pel square
90 that is to be copied is located, for example 25½
pels below and 23 pels to the right of the macroblock
to be reconstructed.
With the above discussion in mind, we find that the Vetro,
Ng and Bose combination teaches the claimed invention. First,
Vetro substantially teaches the limitations set forth in
representative claim 1. Particularly, we find that Vetro teaches
a frequency domain down-conversion of HDTV using a low-resolution
motion compensation (MC) scheme that includes specifying high
12
Page: Previous 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Next
Last modified: November 3, 2007