Title: Internet Video Protocols
1Internet Video Protocols Applications
2003. 5. 27
2Contents
- Introduction
- Building loss tolerant video delivery system
- Handling Delay mechanism
- Congestion Control
- RTP RTCP
- Case Study SALSA
- Conclusion References
3Introduction
- Video delivery applications on Internet
- Video as a part of Information-on-demand(IoD)
- Applications involved transport of live video
- Bi-directional interactive video applications
- Video highest tolerance to packet losses
- To make the best use of this loss tolerance in
dealing with the delivery impairments - Important impairments of packet delivery over
internet - Loss
- Delay
4Loss Tolerant Video Delivery System
- Application Layer Framing (ALF)
- Constrain the effect of a packet loss to a single
packet - Increase error resilience
- Fragmentation smaller than Max. Transmission Unit
(MTU) - Repetition of higher layer headers within each
packet - Reference Picture Selection
- Proper Key Frame Updates
- Increasing key frame rate ? reduce waiting time
for playback - Slices or MBs instead of complete frames
- Layered coding
- Useful for addressing the heterogeneity issue
- Not only video information is contained in base
layer
5Handling Delay Mechanism
- Propagation delay on the wires
- Queuing delay on the routers
- Play-out buffers size management
- Determine the initial minimum buffer size
- Implement a mechanism to change the buffer size
in real-time with minimal effect on the quality
of the playback - Effective sampling rate modification based on
frame repeat/drop - Global synchronization will be need
- Synchronous delivery channels with low loss rate
- Not suitable for multicast of heterogeneous
receivers
6Congestion Control
- Bottleneck bandwidth
- Upper limit on the speed of delivering data from
end to end - Obstacles of rate control
- Busty nature of the Internet traffic
- Multicast applications
- Rate adjustment algorithm based on a combination
of the RTCP reports from all receivers - Sending different layers of the output of a
layered codec to different multicast groups - Rate control of stored video
- Require real-time transcoding
- ? Increase the computational complexity
- Storing the same video encoded at several rates
7RTP RTCP
- Payload type identification
- Special field at the packet header
- Packet sequence numbering
- Packet loss detection, packet re-ordering
- Randomly selected initial sequence number
- Time Stamping (32 bit)
- Encoder/decoder clock matching
- Synchronization of several sources
- Measuring packet arrival jitter
- Source identification
- Synchronization SouRCe identifier (SSRC)
- RTCP
- Periodic transmission of control packets from
participants of a session
8RTP RTCP (Cont.)
- Feedback on the quality of distribution and
timing - RTP reports among sender and receivers
- Fraction of the lost RTP packets since the last
report - Cumulative of packets lost since the beginning
of reception - Packet inter-arrival jitter
- Delay since receiving the last senders report
- Not defined explicit flow control of RTP
- Capable of generating high traffic rates causing
network congestion - Receiver specific feedback for error recovery
- Periodic RTCP
- Contain 64bit Network Time Protocol (NTP)
timestamps - Intra or inter-media synchronization with
synchronized NTP timestamps
9RTP RTCP (Cont.)
- Participant Identification
- Connection between the real identification of an
RTP source - Using Canonical name (CNAME) current SSRC
- Scale the control packet transmission
- Less than 5 of the bandwidth allocated for a
session
10SALSA (System for Audio/visual Live Services and
Applications)
- Two way interactive video over cable TV
- Many applications of bi-directional video
- Personal, tutoring, shopping, etc.
- Layered coding to protect the video from packet
losses - Temporal scalability
- QoS guarantee between end points
- Cable Modem Termination System (CMTS) located at
the cable head-end - Data-over-Cable Service Interface Specification
(DOCSIS) 1.1 - H.263 G.728 ? transmitting base layer over QoS
channel - Usable video under a 30 packet loss rate on the
best effort channel
11Summary Conclusion
- Outline the Internet packet video delivery
- Review some possible remedies to address
feasibility of using the Internet video - To design a cost effective video delivery system
for the Internet is needed - QoS guaranteed network infrastructure for
interactive applications - Associated price structure
- Familiarity with the transport protocol is
essential for video codec designers
12References
- M. R. Civanlar, "Internet video - Protocols
applications," in Proc. Packet Video Workshop
2001, 2001
13Real-Time Internet Video using resilient
Scalable Compression and TCP-Friendly Transport
Protocol
2003. 5. 27
14Contents
- Introduction
- Error-resilient bandwidth-scalable video
compression - TCP-friendly transport protocol
- Conclusion
- references
15Introduction
- Discrete Wavelet Transform (Subband Coding)
- DC subband contains most of the energy
- Generate several layers ? various spatial and
quality resolutions - Multi-priority system
downsample
upsample
?2
?2
H0(z)
G0(z)
X(z)
Y(z)
?
?2
?2
H1(z)
G1(z)
Coding transmission
16Hierarchical block coding (HBC)
- Significance Map
- Binary map indicating whether each coeff. is
significant at the current bit-plane - 44 ? 22 ? indicating each bit
17Bandwidth Scalable Packetization
- Each component contains coeff. Blocks which are
independently compressed using progressive
quantization by hierarchical block coding - Successive bit-planes of a coeff. Block are
intercoded to produce small codewords - ? Fine granularity for layered packetization
18Encoding procedure
- Every time the encoder finishes subband analysis,
it computes and parametized the energy
distribution of the subbands into 8 parameters
compactly transmitted per packet
19Two source of latency in TCP
- Backlog of data when throughput temporarily drops
below the data generating rate - ? filtering a scalable video bit-stream to meet
the instantaneous throughput - Retransmission delay
- ? not performing retransmission
20Throughput estimation of TCP
- On a lightly loaded network,
- TCP sending window ? amount of buffer space (B)
- ? Calculating average throughput (T) with
buffer space - TCP throughput to the packet loss rate (p)
- (kconstant, MSS Max. segment size)
21Throughput estimation of TCP (Cont.)
- Packet losses within one RTT are considered as a
single congestion event, then - C of congestion events in an observation
window - D amount of data in units of MSS (Max. Segment
Size) transmitted in the same observation window - Throughput estimated by (1) and (2) will show
large variability - ? Throughput is calculated using an additional
RTT estimate that is measured using an accurate
clock
22TCP Friendly Transport Protocol
- Modulated by flow controller not to exceed
throughput given by (1) and (2) - If source attempts to make a fixed of
transmissions per second - If the size of each transmission
- Reduce transmission overhead by combining
consecutive small transmission
23Performance comparison
(proposed)
(T) compared
(T) scalable compression scheme producing
linearly dependent packet
24FEC Packet Replication
(P) Ave. MSE 100.53
(T) Ave. MSE 201
- Higher ave. distortion of scheme (T) compared
with scheme (P) with or without FEC
25Conclusion
- Low latency video transmission scheme consisting
of a TCP-friendly flow controller - By elimination of buffering in flow control
- Overall latency propagation delay decoding
time - Bandwidth-scalable compression scheme producing
individually decodable packets - Produce relatively constant video quality in face
of packet losses as compared to MPEG and other
schemes - Independently decoded packets reduces the
resulting distortion even after an adaptive FEC
scheme is applied to protect the non-resilient
bitstream
26References
- Wai-tian Tan and Avideh Zakhor. "Real-Time
Internet Video Using Error Resilient Scalable
Compression and Tcp-Friendly Transport Protocol",
in IEEE Trans. on Multimedia, 1999