Information Technology Reference
In-Depth Information
high-spatial-resolution video is not available in the raw format and the users watch
the sequence asynchronously, the high-spatial-resolution video has to be decoded
before cropping the RoI sequence for each user.
Spatial-random-access-enabled video compression limits the load of encoding to
a one-time encoding of the video, possibly with multiple resolution layers to support
continuous zoom. This is beneficial for streaming both live content as well as pre-
stored content to multiple users. Even when the video is played back from locally
stored media, a different RoI trajectory has to be accommodated each time a user
watches the content. Spatial random access allows the video player to selectively de-
code relevant regions only. This chapter presents a spatial-random-access-enabled
video coding scheme in detail that allows the receiver to start decoding a new re-
gion, with an arbitrary zoom factor, during any frame-interval instead of having to
wait for the end of a GOP or having to transmit extra slices from previous frames.
Background extraction can be used with such a coding scheme to reduce transmis-
sion bit-rate as well as the size of the stored video. We also show how to choose the
slice size to attain the right balance between storage and mean transmission bit-rate.
Pre-fetching helps to reduce the latency of interaction. Irrespective of whether
pre-fetching is employed or not, having a base layer helps render missing parts of
the RoI. This way, the system can always render the RoI chosen by the user, thus
offering accurate and low-latency RoI control. The chapter presents several RoI
prediction techniques for pre-fetching. Some techniques are employed at the client,
some at the server and some are distributed between the server and the client. For
example, the server can send the trajectories of key objects in the video to the clients
to aid their RoI prediction modules.
This chapter also shows how to use P2P streaming to drastically reduce the band-
width required from the server for supporting increasing numbers of users. It is cru-
cial for this approach that the P2P overlay reacts quickly to the changing RoIs of the
peers and limits the disruption due to the changing relationships among the peers.
The IRoI P2P protocol presented in this chapter makes sure that a child-peer experi-
ences smooth transition from old parent to new parent when the old parent willfully
unsubscribes a multicast tree that is no longer required for its RoI. Typically, when
users choose regions from a high-spatial-resolution video, some regions are more
popular than others. It is very important, especially when the server has limited
bandwidth, to judiciously allocate the available rate among the regions streamed by
the server.
Spatial-random-access-enabled video coding plays an important role in the P2P
distribution system. It simplifies the peer's task of choosing which multicast trees
to join on-the-fly. A scheme based on multiple representations coding of the slices
might further reduce the download rate required by each peer. However, such a
scheme might reduce the degree of overlaps and affect the gains possible from the
P2P approach apart from requiring more storage space at the server. The IRoI P2P
system presented in this chapter assumes that peers watch the video synchronously.
If peers can watch any time-segment, i.e., rewind and fast-forward then the rele-
vant data could still be retrieved from each others' cache. Since storage is becom-
ing cheaper, the cache size employed by the peers for storing previously received
