Graphics Reference
In-Depth Information
The elementary operation of the angular prediction described in Sect.
4.3.1.3
can
be simplified as:
p
D
.
u
a
C
v
b
C
16/ >> 5
(4.27)
where
u
and
v
are two pre-computed constants, and
a
and
b
are two pre-fetched
reference sample values. It suggests that the angular prediction requires a total of
five operations including two multiplications, two additions and one shift per one
predicted sample. In the H.264/AVC case, the directional prediction process can be
expressed as:
p
D
.a
C
2
b
C
c
C
2/ >> 2
(4.28)
where
a
,
b
and
c
are three pre-fetched reference sample values. It means that the
H.264/AVC process requires also a total of five operations including three additions
and two shifts.
The directional prediction of H.264/AVC can be implemented by using only
multiplication-free operations easily, thus it is reported to have less computational
complexity compared with the angular prediction of HEVC in some specific
architecture, but the difference tends to be relatively minor [
8
]. On the other hand,
all the HEVC angular prediction modes use the same prediction process regardless
of the block sizes whereas H.264/AVC defines slightly different operations for 4
4,
8
8 and 16
16 block sizes. This property of HEVC can be considered very useful
when it comes to implementing the HEVC prediction modes as there is such a
large amount of different combinations of prediction block sizes and the prediction
directions in HEVC.
In [
8
], a more detailed decoder complexity analysis has been carried out by using
decoding times measured in an optimized software implementation [
2
]. For the
detailed information such as the relative decoding time of intra prediction to overall
decoding time or cycle counts for each prediction mode, please refer to [
8
].
4.7
Main Still Picture Profile and Its Applications
HEVC version 1 includes a specific profile for still picture coding purposes [
15
]. The
“Main Still Picture profile” defines bitstream characteristics and decoder operation
for using the HEVC toolset to efficiently represent both typical camera captured
still pictures as well as synthetic images. Typical use cases for the profile include
traditional photography, capturing still pictures during video recording, extraction
of still pictures from video sequences, sharing pictures over Internet or other
communication channels, and different computational photography applications.
HEVC Still Picture profile is able to provide a very bandwidth efficient way to
represent high quality images in each of these use cases. In the future extensions
