EXFO EXpert IPTV Test Tools (FTB-200v2) - Manual (Page 49)

49
As a frame often spans multiple packets, and a typical video stream includes interpolated
frames (P‐frames and B‐frames), a given packet loss rate can result in a frame loss rate six
times higher [9]. See Figure 2‐1 above.
In order to accommodate IPTV transmission network delay and low levels of delay variation,
a playout buffer is used to temporarily store incoming frames. For streaming video, such as
DVB or IPTV, it is permissible to apply arbitrary delays, and hence the playout buffer can be
quite large. Interactive videoconferencing requires a relatively low delay, since a participant
needs to respond immediately to questions or opinions of the other participants. Packets
arriving too late may be discarded and will appear as lost packets.
Traditionally, video quality is measured by comparing the video sequence that has been
processed by a video system to the original reference video sequence. It is not feasible to
implement a mid‐stream real‐time monitoring system using this approach. This is because it
needs to directly access the video source and decode video content, and has high
computational complexity.
To overcome these disadvantages, it is therefore desirable to monitor transmission
impairments in order to ensure good video quality. VQmon/HD is a non‐intrusive monitoring
technology that can passively measure the characteristics of live packet‐based video streams
and report quality scores in real‐time. VQmon/HD models the way that time‐varying
impairments, most notably burst packet loss and possible playout buffer discards, affect
perceived video quality.
A.1.1.1 Impact of Packet Loss and Jitter on Video Quality
Common video codecs, such as MPEG and H.263 implementations, use a combination
of intra‐ and inter‐frame coding methods. For intra‐frame encoding (I frame) the image frame
is divided into blocks, a Discrete Cosine Transform is used to convert each block to a set of
coefficients and then variable length coding is applied. A group of blocks are combined into a
single entity (slice), which can be carried within a single packet. If a transmission error occurs
then the whole group may be lost, creating a “stripe” within the decoded image. For example,
because the DC coefficients within each block are predicatively encoded from the first block in
the slice, an error makes this information unusable for the remainder of the slice. Some errors
may damage the frame structure and render the whole frame unusable.