Graphics Reference
In-Depth Information
Fig. 6.8
Construction of
default quantization matrices
for transform block sizes
16
16x16
32 by using
the default quantization
matrix of size 8
16 and 32
8
8x8
32x32
Tabl e 6. 4
Quantization group size for different coding tree unit sizes
Quantization
group size for
64
Quantization
group size for
32
Quantization
group size for
16
diff_cu_qp_delta_depth
64 CTU
32 CTU
16 CTU
0
64
64
32
32
16
16
1
32
32
16
16
8
8
2
16
16
8
8
-
3
8
8
-
-
decoder as shown in Fig.
6.8
) and the quantization matrix entry at the DC
(zero-frequency) position are transmitted. HEVC also allows for prediction of a
quantization matrix from another quantization matrix of the same size. The use
of quantization matrix (termed as scaling matrix in HEVC) is enabled by setting
the flag
scaling_list_enabled_flag
in SPS. When this flag is enabled,
additional flags in SPS and PPS control whether the default quantization matrices
or non-default quantization matrices are used.
6.3.2
QP Parameter Derivation
The quantization step size (and therefore the
QP
value) may need to be changed
within a picture for e.g. rate control and perceptual quantization purposes. HEVC
allows for transmission of a delta
QP
value at a
quantization group
(QG) level to
allow for
QP
changes within a picture. This is similar to H.264/AVC that allows
for modification of QP values at a macroblock level. The QG size is a multiple of
coding unit size that can vary from 8
8to64
64 depending on the coding tree
unit (CTU) size and the syntax element
diff_cu_qp_delta_depth
as shown
in Table
6.4
.
