Graphics Reference
In-Depth Information
prediction modes of luma blocks to less than half. With the hardware design, two
metrics need to be better balanced for a higher throughout which represnets the clock
cycles and the critical path.
The essay is organized as the follows: the implemented fast mode decision algo-
rithm is proposed in section 2; the intra prediction hardware architecture is explained
in Section 3; the synthesis results are presented in Section 4; and Section 5 presents
the conclusion.
2
Implemented Fast Mode Decision Algorithm of Intra
Prediction
This algorithm gives a fast way of calculating the best prediction mode of intra pre-
diction, and which is proposed based on the inherent correlation between the intra
prediction modes and blocks. In this way, the candidate prediction modes can reduce
to less than half.
2.1
The Fast Decision of Partitioning in H.264 Intra Prediction
As mentioned above, the 4×4 intra prediction is preferred in the macroblock which
has high texture complexity, otherwise, the 16×16 intra prediction is been chosen.
Therefore, we can determine the partition by using the calculated texture complexity
of a macroblock before intra prediction coding. The texture complexity of luma
blocks on the directions of vertical, horizontal and oblique 135˚ are defined in the
following way:
1
m
m
−
1
.
=
(1)
TC
=
P
+
−
P
vertical
i
,
j
1
i
,
j
m
(
m
−
1
i
0
j
=
0
m
m
−
1
1
.
=
TC
=
P
−
P
(2)
horizontal
i
+
1
j
i
,
j
m
(
m
−
1
j
0
i
=
0
1
m
−
1
m
.
(3)
TC
=
P
−
−
P
angle
i
1
j
1
i
,
j
(
m
−
1
)(
m
−
1
j
==
00
i
where m is the size of luma block, P
i,j
is the original pixel values located in the i
row and j column of the luma blocks.
The mean value of the texture complexity is computed as:
1
.
(4)
M
=
(
TC
+
TC
+
TC
)
TC
vertical
horizontal
angle
3
The threshold of texture complexity is computed as:
1
.
TC
=
(
TC
−
M
+
TC
−
M
+
TC
−
M
)
(5)
th
vertical
TC
horizontal
TC
angle
TC
M
TC
Search WWH ::
Custom Search