Game Development Reference
In-Depth Information
Table 9.3
Coding
performance of the divisive
normalization scheme in
RD4.0 (in terms of SSIM)
Sequence
RA (%)
LD (%)
ClassA
−
2.5
−
4.4
ClassB
−
2.4
−
3.8
ClassC
−
6.1
−
8.1
ClassD
−
10.1
−
14.30
ClassE
−
4.5
−
7.6
Average
−
5.1
−
7.6
mean value have effectively larger quantization step and vice versa. By doing so, we
are borrowing bits from the regions which are perceptually less important and using
them for the regions with more perceptual relevance, as far as SSIM is concerned,
so that all the regions in the frame conceptually have uniform perceptual distortion.
It is important to note that the reference point, mean AC and DC energies, is highly
dependent on the content of the video frame. The frames with significant texture
regions are likely to get more perceptual improvement because the texture regions
are the main beneficiaries of the spatially adaptive normalization process.
The divisive normalization scheme (Wang et al.
2013
) is performed in AVS2 stan-
dard, where the transform coefficients are quantized with the corresponding divisive
normalization factor and the input quantization level. The experimental results are
shown in Table
9.3
. It is observed that significant coding performance is achieved in
terms of SSIM, which demonstrates the superior performance of the divisive nor-
malization scheme.
9.2 Video Coding Application System
The output of a video or audio coder is called an Elementary stream, which is not
suitable for storage and transport. For convenience, the metadata need to be divided
into small data blocks in a Packetized Elementary Stream (PES). Then, the metadata
in PES, video, audio and auxiliary data, are multiplexed into Program Stream (PS)
or Transport Stream (TS) for storage and transport, e.g., Fig.
9.5
(Haskell
1997
).
Since the packetizing disrupts the time axis in video and audio data, the time
stamps must be included in PES to identify the order of these data blocks. The two
types of multiplexed stream, PS and TS, are designed for different applications. The
program streams are one way of combining several PES packet streams, which are
encoded using the same master clock or system time clock (STC) and are more
suitable for storage applications, such as Blu-ray Disk. The transport streams can
combine one or more programs with the same or different STC into one single stream.
The transport streams are designed for less reliable transmission, i.e., terrestrial or
satellite broadcast. Further, a transport stream may carry multiple programs. But
there are not strict hierarchical structures for the two types of stream, because they are
