Graphics Reference
In-Depth Information
signaling the selected modes, but decrease the resulting average rate-distortion cost
for coding the prediction residuals, presuming a suitable encoder decision algorithm.
One set
of partitioning modes improves the coding efficiency relative to another
set if it yields a smaller expectation value,
E
A
D.p/
C
R.p/
;
min
8p2
A
(3.2)
of the Lagrangian rate-distortion cost for typical video content. It should, however,
be noted that for a larger set of possible subdivision modes, in general, an encoder
also requires a higher computational complexity for evaluating the set of supported
modes. Hence, when designing a standard, a reasonable compromise between the
potential coding efficiency and the required encoder complexity has to be chosen.
Due to continuous improvements in computing power, newer video coding stan-
dards support an increased set of coding options. In the development of HEVC, it
was further taken into account that the coding of high and ultra-high definition video
becomes more and more important. For dealing with such high resolutions, it is
generally advantageous to support larger block sizes for both motion-compensated
prediction and transform coding. But for adapting the block partitioning to the
local properties of pictures, it is also important to additionally support small block
sizes. Both objectives have been addressed in HEVC by introducing a hierarchical
block partitioning concept based on a simple and unified but yet efficient quadtree
syntax [
27
,
44
]. In addition, this quadtree-based block partitioning concept allows
the application of fast optimal tree pruning algorithms [
5
] in the encoder for
determining the best block partitioning in terms of Lagrangian rate-distortion cost.
3.2.1
Coding Tree Blocks and Coding Tree Units
In all prior video coding standards [
11
,
12
,
14
-
17
] of the ITU-T and ISO/IEC, each
picture of a video sequence is partitioned into so-called macroblocks. A macroblock
consists of a 16
16 block of luma samples and, in the 4:2:0 chroma sampling
format, two associated 8
8 blocks of chroma samples (one for each chroma
component). The macroblocks can be considered as the basic processing units
in these standards. For each macroblock of a picture, a coding mode has to be
selected by the encoder. The chosen macroblock coding mode determines whether
all samples of a macroblock are predicted using intra-picture prediction or motion-
compensated prediction. Depending on the features supported in the actual standard,
it can additionally determine the partitioning of the macroblock into subblocks
that are used for motion-compensated prediction or intra-picture coding. The
macroblock size of 16
16 luma samples represents the largest block size that can
be used for signaling prediction parameters such as motion data.
Although video coding standards such as H.262
j
MPEG-2 Video [
16
]
and H.264
j
MPEG-4 AVC [
17
] are used today for storing and transmitting
