Information Technology Reference
In-Depth Information
experimented by users in phone calls. This model (E-model) is based on the prem-
ise that: "Psychological factors on the psychological scale are additive". The E-
model takes into account factors such as packet loss, delay, and others like the
equipment impairment and the packet loss robustness factors that depends on the
codec used in the connection, as described by the recommendation ITU-T G.113.
When we are dealing with video streams, there are multiple parameters that can
change between two videos even if they are coded with the same codec. Most of
up-to-date codecs define different profiles with several resolutions, frames per
second, bit rates, etc. Some approaches have used automatic measures of the video
quality based on the Peak Signal to Noise Ratio (PSNR) [25]. However, it is im-
possible to have the original picture at the destination and, therefore, the PSNR
can only be used to extract some parameters of the Mean Opinion Square (MOS)
in function of the packet loss.
If we analyze the PSNR picture by picture, it does not take into account the
transition between pictures, and the human perception is very sensitive to this
transitions. Thus, this method does not give a good perceptual approach of a video
sequence. Other solutions such as the Video Quality Metric (VQM) offer an ob-
jective quality approach to help in the design and control of digital transmission
systems [26]. In [27] appears a study which takes into account the perceptual im-
pression of packet loss, variation in the frame rate and synchronization with the
audio signal. Finally, a more sophisticated study considering the opinion of the us-
ers is also explained in [25].
4.3 An Interoperable QoS Model for Video Transmission
Exploiting Cross-Layer Interactions
4.3.1 QoS Probes and Mapping
In our study we made an analysis of the perceptual quality for video transmission
with different ranges of parameters. The video sequences have encoded using
AVC/H.264 with a frame rate from 6.25 to 25 frames per second. The video
bandwidth used is between 150 and 1500 kbps, and finally, we introduced simu-
lated random packet loss up to 10%. With all this ranges we build a huge reposi-
tory of videos with some of their parameters modified. In this way, we can
observe not only the effect on the subjective quality of the video (QoE) when
varying only one parameter, but also can simultaneously study the cumulative ef-
fect on the quality of several of them (QoS).
A public survey [28] has been distributed in order to include as wide and hetero-
geneous audience as possible in both, internal and external approach, in a national
and European environment. Each person watched a minimum of 10 videos randomly
selected from the repository and rated the quality of the video between a value of 1
for a bad quality and 5 for a perfect quality. From this evaluation and the corre-
sponding content parameters, we were able to derive the formulas as presented in the
subsequent sections (bottom-up) to be used in a video transmission system.
