Compression - 3DTV: Processing and Transmission of 3D Video Signals - page 71

CAVLC is computationally less intensive and simpler than CABAC, where
CABAC generally saves more bits during encoding when compared to
CAVLC with its 'non-integer length' code-word assignment capability. Both
methods provide a significant increase in the compression with respect to the
entropy coding methods used in former video coding standards. Huffman
coding is one example of the variable length entropy coding techniques used
in former coding standards.

3.2.2 HighEfficiencyVideoCoding(HEVC)

Recent standardization efforts have yielded the new High Efficiency Video
Coding (HEVC) (also commonly referred to as H.265, or MPEG-H Part 2),
which is expected to be more efficient than its predecessor AVC. Similar to
the AVC standard, HEVC is also a project of the Joint Collaborative Team
on Video Coding (JCT-VC) of ISO/IEC-MPEG and ITU-VCEG. HEVC aims
to provide significantly improved compression efficiency compared to the
AVC High Profile. The intention is to reduce bit-rate requirements by up to
50% with comparable image quality [6]. The committee draft standard was
published in February 2012, while the final international draft standard was
released in January 2013. This section provides a brief overview of some
of the new features in HEVC, and further details concerning HEVC can be
found in [7].
One of the prominent features of the HEVC model is the use of variable
block sizes in a tree structure, ranging from 8 × 8to64 × 64, unlike AVC

that uses fixed size macroblocks of size 16 16. The largest blocks are called

Coding Tree Blocks (CTBs), whose size may be changed from sequence to
sequence. CTBs may be specified as being 16

×

64. CTBs
may be further segmented into Coding Units (CUs). CUs' sizes range from
8

× 16, 32 × 32, or 64 ×

64. An example of how a CTB may be split into different CU
sizes is shown in Figure 3.3.
Coding Units can be segmented into Prediction Units (PUs), where each
segment of a PU may be given a different Motion Vector. The Inter PU types
specified in HEVC are shown in Figure 3.4. The bottom half of Figure 3.4
shows the partitions that may be obtained using the Asymmetric Motion Par-
titioning (AMP) option, which divides a block into two unequal parts (a three
quarters - one quarter split). Two new tools are employed to make motion
representation more efficient: Advanced Motion Vector Prediction (AMVP)
and Merge mode. AMVP allows more efficiency prediction and coding of
motion vectors, by allowing a predicted motion vector to be formed from
a list of candidate prediction motion vectors. Candidates are derived from
temporally co-located PUs in other frames, and from spatially adjacent PUs.
Merge mode allows motion vectors to be predicted by using an index
value to select from particular motion vector predictors. This allows the
decoder to predict and derive the motion vectors for Merge mode coded
PUs. Together these tools make it possible to describe more detailed motion

× 8to64 ×