Graphics Reference
In-Depth Information
a
b
Linear interpolation
B
C
C
Y
D
Y
Block to be
predicted
Block to be
predicted
B
X
A
X
A
Fig. 4.4
Two types of filtering process of reference samples: (
a
) shows a three-tap filtering using
two neighboring reference samples. Reference sample
X
is replaced by the filtered value using
A
,
X
and
B
while reference sample
Y
is replaced by the filtered value using
C
,
Y
and
D
.(
b
)Showsa
strong intra smoothing process using corner reference samples. Reference sample
X
is replaced by
a linearly filtered value using
A
and
B
while
Y
is replaced by a linearly filtered value using
B
and
C
j
pŒ
1Œ
1
C
pŒ
1Œ2N
1
2p Œ
1ŒN
1
j
<.1<<.b
5// (4.5)
where
b
specifies the sample bit depth. If the above two inequalities are satisfied
and the prediction block size is equal to 32
32, the non-corner reference samples
of
p
[
1][0], :::,
p
[
1][62] and
p
[0][
1], :::,
p
[62][
1] are generated as:
pŒ
1Œy
D
..63
y/
pŒ
1Œ
1
C
.y
C
1/
pŒ
1 Π63
C
32/ >> 6
(4.6)
pŒx Œ
1
D
..63
x/
pŒ
1Œ
1
C
.x
C
1/
pŒ63Œ
1
C
32/ >> 6
(4.7)
for
y
D
0 :::62 and
x
D
0 :::62.
This process is referred to as strong intra smoothing as it substitutes nearly all
the original reference samples with interpolated ones. The process was developed
to remove some blocking and contouring artifacts visible on extremely smooth
image areas and it can be selectively turned on or off by the syntax element
strong_intra_smoothing_enabled_flag
in a sequence parameter set. Figure
4.4
illustrates the two types of filtering process of reference samples.
4.3
Intra Sample Prediction
In order to predict different kinds of content efficiently HEVC supports a range
of sample prediction methods. The angular intra prediction is designed to model
different directional structures typically present in pictures. Whereas planar and DC
