Graphics Reference
In-Depth Information
M
×
M
M
×(
M
/2)
(
M
/2) ×
M
(
M
/2) × (
M
/2)
M
×(
M
/4)
M
×(3
M
/4)
(
M
/4) ×
M
(3
M
/4) ×
M
Fig. 3.6
Supported partitioning modes for splitting a coding unit (CU) into one, two, or four
prediction units (PU). The .M=2/
.M=2/ mode and the modes shown in the
bottom row
are
not supported for all CU sizes
the average coding efficiency of transform coding (in terms of the average mean
squared error for a given bit rate) typically increases with the transform size [
1
,
42
].
Hence, the syntax feature that allows the splitting of a coding block (or a block used
for signaling the intra prediction mode) into multiple transform blocks provides the
possibility to select a suitable trade-off between the intra prediction and transform
coding efficiency for the considered block. From a different point of view, it can
also be argued that coding a single intra prediction mode for multiple transform
blocks (i.e., the blocks that are actually used for intra prediction) represents a way
for reducing the bit rate required for transmitting the intra prediction modes.
If a CU is coded using inter-picture prediction, the luma and chroma CBs can be
further split into so-called prediction blocks. A prediction block (PB) is a block of
samples of the luma or a chroma component that uses the same motion parameters
for motion-compensated prediction. The motion parameters include the number of
motion hypotheses (which is either one or two) as well as the reference picture
index and motion vector for each of the motion hypotheses. For both chroma CBs
of a CU, the same splitting as for the luma CB is used. The luma PB and chroma
PBs, together with the associated syntax, form a prediction unit (PU). For each PU,
a single set of motion parameters is signaled in the bitstream, which is used for
motion-compensated prediction of the luma PB and the chroma PBs.
HEVC supports eight different modes for partitioning a CU into PUs. As
illustrated in Fig.
3.6
, a CU can either be coded as a single PU or it can be split
into two or four rectangular PUs. The partitioning mode in which the entire CU
is coded as a single PU is referred to as M
M mode. If a CU is split into
four PUs, the resulting PUs represent square blocks of the same size and the
partitioning mode is referred to as .M=2/
.M=2/ mode. Since the splitting of
a CU into four equally-sized square PUs is conceptually equivalent to splitting
the corresponding picture block into four CUs and coding each of these CUs as
