| 1 |
<pre>Network Working Group T. Turletti |
| 2 |
Request for Comments: 2032 MIT |
| 3 |
Category: Standards Track C. Huitema |
| 4 |
Bellcore |
| 5 |
October 1996 |
| 6 |
|
| 7 |
|
| 8 |
<span class="h1">RTP Payload Format for H.261 Video Streams</span> |
| 9 |
|
| 10 |
Status of this Memo |
| 11 |
|
| 12 |
This document specifies an Internet standards track protocol for the |
| 13 |
Internet community, and requests discussion and suggestions for |
| 14 |
improvements. Please refer to the current edition of the "Internet |
| 15 |
Official Protocol Standards" (STD 1) for the standardization state |
| 16 |
and status of this protocol. Distribution of this memo is unlimited. |
| 17 |
|
| 18 |
Table of Contents |
| 19 |
|
| 20 |
<a href="#section-1">1</a>. Abstract ............................................. <a href="#page-1">1</a> |
| 21 |
<a href="#section-2">2</a>. Purpose of this document ............................. <a href="#page-2">2</a> |
| 22 |
<a href="#section-3">3</a>. Structure of the packet stream ....................... <a href="#page-2">2</a> |
| 23 |
<a href="#section-3.1">3.1</a> Overview of the ITU-T recommendation H.261 .......... <a href="#page-2">2</a> |
| 24 |
<a href="#section-3.2">3.2</a> Considerations for packetization .................... <a href="#page-3">3</a> |
| 25 |
<a href="#section-4">4</a>. Specification of the packetization scheme ............ <a href="#page-4">4</a> |
| 26 |
<a href="#section-4.1">4.1</a> Usage of RTP ........................................ <a href="#page-4">4</a> |
| 27 |
<a href="#section-4.2">4.2</a> Recommendations for operation with hardware codecs .. <a href="#page-6">6</a> |
| 28 |
<a href="#section-5">5</a>. Packet loss issues ................................... <a href="#page-7">7</a> |
| 29 |
<a href="#section-5.1">5.1</a> Use of optional H.261-specific control packets ...... <a href="#page-8">8</a> |
| 30 |
<a href="#section-5.2">5.2</a> H.261 control packets definition .................... <a href="#page-9">9</a> |
| 31 |
<a href="#section-5.2.1">5.2.1</a> Full INTRA-frame Request (FIR) packet ............. <a href="#page-9">9</a> |
| 32 |
<a href="#section-5.2.2">5.2.2</a> Negative ACKnowledgements (NACK) packet ........... <a href="#page-9">9</a> |
| 33 |
<a href="#section-6">6</a>. Security Considerations .............................. <a href="#page-10">10</a> |
| 34 |
Authors' Addresses ..................................... <a href="#page-10">10</a> |
| 35 |
Acknowledgements ....................................... <a href="#page-10">10</a> |
| 36 |
References ............................................. <a href="#page-11">11</a> |
| 37 |
|
| 38 |
<span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. Abstract</span> |
| 39 |
|
| 40 |
This memo describes a scheme to packetize an H.261 video stream for |
| 41 |
transport using the Real-time Transport Protocol, RTP, with any of |
| 42 |
the underlying protocols that carry RTP. |
| 43 |
|
| 44 |
This specification is a product of the Audio/Video Transport working |
| 45 |
group within the Internet Engineering Task Force. Comments are |
| 46 |
solicited and should be addressed to the working group's mailing list |
| 47 |
at [email protected] and/or the authors. |
| 48 |
|
| 49 |
|
| 50 |
|
| 51 |
|
| 52 |
<span class="grey">Turletti & Huitema Standards Track [Page 1]</span></pre> |
| 53 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span> |
| 54 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 55 |
|
| 56 |
|
| 57 |
<span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. Purpose of this document</span> |
| 58 |
|
| 59 |
The ITU-T recommendation H.261 [<a href="#ref-6">6</a>] specifies the encodings used by |
| 60 |
ITU-T compliant video-conference codecs. Although these encodings |
| 61 |
were originally specified for fixed data rate ISDN circuits, |
| 62 |
experiments [<a href="#ref-3" title="H.261 software codec for videoconferencing over the Internet INRIA Research Report no 1834">3</a>],[<a href="#ref-8" title="T. Turletti">8</a>] have shown that they can also be used over |
| 63 |
packet-switched networks such as the Internet. |
| 64 |
|
| 65 |
The purpose of this memo is to specify the RTP payload format for |
| 66 |
encapsulating H.261 video streams in RTP [<a href="#ref-1" title=""RTP: A Transport Protocol for Real-Time Applications"">1</a>]. |
| 67 |
|
| 68 |
<span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>. Structure of the packet stream</span> |
| 69 |
|
| 70 |
<span class="h3"><a class="selflink" id="section-3.1" href="#section-3.1">3.1</a>. Overview of the ITU-T recommendation H.261</span> |
| 71 |
|
| 72 |
The H.261 coding is organized as a hierarchy of groupings. The video |
| 73 |
stream is composed of a sequence of images, or frames, which are |
| 74 |
themselves organized as a set of Groups of Blocks (GOB). Note that |
| 75 |
H.261 "pictures" are referred as "frames" in this document. Each GOB |
| 76 |
holds a set of 3 lines of 11 macro blocks (MB). Each MB carries |
| 77 |
information on a group of 16x16 pixels: luminance information is |
| 78 |
specified for 4 blocks of 8x8 pixels, while chrominance information |
| 79 |
is given by two "red" and "blue" color difference components at a |
| 80 |
resolution of only 8x8 pixels. These components and the codes |
| 81 |
representing their sampled values are as defined in the ITU-R |
| 82 |
Recommendation 601 [<a href="#ref-7">7</a>]. |
| 83 |
|
| 84 |
This grouping is used to specify information at each level of the |
| 85 |
hierarchy: |
| 86 |
|
| 87 |
- At the frame level, one specifies information such as the |
| 88 |
delay from the previous frame, the image format, and |
| 89 |
various indicators. |
| 90 |
|
| 91 |
- At the GOB level, one specifies the GOB number and the |
| 92 |
default quantifier that will be used for the MBs. |
| 93 |
|
| 94 |
- At the MB level, one specifies which blocks are present |
| 95 |
and which did not change, and optionally a quantifier and |
| 96 |
motion vectors. |
| 97 |
|
| 98 |
Blocks which have changed are encoded by computing the discrete |
| 99 |
cosine transform (DCT) of their coefficients, which are then |
| 100 |
quantized and Huffman encoded (Variable Length Codes). |
| 101 |
|
| 102 |
The H.261 Huffman encoding includes a special "GOB start" pattern, |
| 103 |
composed of 15 zeroes followed by a single 1, that cannot be imitated |
| 104 |
by any other code words. This pattern is included at the beginning of |
| 105 |
|
| 106 |
|
| 107 |
|
| 108 |
<span class="grey">Turletti & Huitema Standards Track [Page 2]</span></pre> |
| 109 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span> |
| 110 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 111 |
|
| 112 |
|
| 113 |
each GOB header (and also at the beginning of each frame header) to |
| 114 |
mark the separation between two GOBs, and is in fact used as an |
| 115 |
indicator that the current GOB is terminated. The encoding also |
| 116 |
includes a stuffing pattern, composed of seven zeroes followed by |
| 117 |
four ones; that stuffing pattern can only be entered between the |
| 118 |
encoding of MBs, or just before the GOB separator. |
| 119 |
|
| 120 |
<span class="h3"><a class="selflink" id="section-3.2" href="#section-3.2">3.2</a>. Considerations for packetization</span> |
| 121 |
|
| 122 |
H.261 codecs designed for operation over ISDN circuits produce a bit |
| 123 |
stream composed of several levels of encoding specified by H.261 and |
| 124 |
companion recommendations. The bits resulting from the Huffman |
| 125 |
encoding are arranged in 512-bit frames, containing 2 bits of |
| 126 |
synchronization, 492 bits of data and 18 bits of error correcting |
| 127 |
code. The 512-bit frames are then interlaced with an audio stream |
| 128 |
and transmitted over px64 kbps circuits according to specification |
| 129 |
H.221 [<a href="#ref-5">5</a>]. |
| 130 |
|
| 131 |
When transmitting over the Internet, we will directly consider the |
| 132 |
output of the Huffman encoding. All the bits produced by the Huffman |
| 133 |
encoding stage will be included in the packet. We will not carry the |
| 134 |
512-bit frames, as protection against bit errors can be obtained by |
| 135 |
other means. Similarly, we will not attempt to multiplex audio and |
| 136 |
video signals in the same packets, as UDP and RTP provide a much more |
| 137 |
efficient way to achieve multiplexing. |
| 138 |
|
| 139 |
Directly transmitting the result of the Huffman encoding over an |
| 140 |
unreliable stream of UDP datagrams would, however, have poor error |
| 141 |
resistance characteristics. The result of the hierachical structure |
| 142 |
of H.261 bit stream is that one needs to receive the information |
| 143 |
present in the frame header to decode the GOBs, as well as the |
| 144 |
information present in the GOB header to decode the MBs. Without |
| 145 |
precautions, this would mean that one has to receive all the packets |
| 146 |
that carry an image in order to properly decode its components. |
| 147 |
|
| 148 |
If each image could be carried in a single packet, this requirement |
| 149 |
would not create a problem. However, a video image or even one GOB by |
| 150 |
itself can sometimes be too large to fit in a single packet. |
| 151 |
Therefore, the MB is taken as the unit of fragmentation. Packets |
| 152 |
must start and end on a MB boundary, i.e. a MB cannot be split across |
| 153 |
multiple packets. Multiple MBs may be carried in a single packet |
| 154 |
when they will fit within the maximal packet size allowed. This |
| 155 |
practice is recommended to reduce the packet send rate and packet |
| 156 |
overhead. |
| 157 |
|
| 158 |
To allow each packet to be processed independently for efficient |
| 159 |
resynchronization in the presence of packet losses, some state |
| 160 |
information from the frame header and GOB header is carried with each |
| 161 |
|
| 162 |
|
| 163 |
|
| 164 |
<span class="grey">Turletti & Huitema Standards Track [Page 3]</span></pre> |
| 165 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span> |
| 166 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 167 |
|
| 168 |
|
| 169 |
packet to allow the MBs in that packet to be decoded. This state |
| 170 |
information includes the GOB number in effect at the start of the |
| 171 |
packet, the macroblock address predictor (i.e. the last MBA encoded |
| 172 |
in the previous packet), the quantizer value in effect prior to the |
| 173 |
start of this packet (GQUANT, MQUANT or zero in case of a beginning |
| 174 |
of GOB) and the reference motion vector data (MVD) for computing the |
| 175 |
true MVDs contained within this packet. The bit stream cannot be |
| 176 |
fragmented between a GOB header and MB 1 of that GOB. |
| 177 |
|
| 178 |
Moreover, since the compressed MB may not fill an integer number of |
| 179 |
octets, the data header contains two three-bit integers, SBIT and |
| 180 |
EBIT, to indicate the number of unused bits in the first and last |
| 181 |
octets of the H.261 data, respectively. |
| 182 |
|
| 183 |
<span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>. Specification of the packetization scheme</span> |
| 184 |
|
| 185 |
<span class="h3"><a class="selflink" id="section-4.1" href="#section-4.1">4.1</a>. Usage of RTP</span> |
| 186 |
|
| 187 |
The H.261 information is carried as payload data within the RTP |
| 188 |
protocol. The following fields of the RTP header are specified: |
| 189 |
|
| 190 |
- The payload type should specify H.261 payload format (see |
| 191 |
the companion RTP profile document <a href="./rfc1890">RFC 1890</a>). |
| 192 |
|
| 193 |
- The RTP timestamp encodes the sampling instant of the |
| 194 |
first video image contained in the RTP data packet. If a |
| 195 |
video image occupies more than one packet, the timestamp |
| 196 |
will be the same on all of those packets. Packets from |
| 197 |
different video images must have different timestamps so |
| 198 |
that frames may be distinguished by the timestamp. For |
| 199 |
H.261 video streams, the RTP timestamp is based on a |
| 200 |
90kHz clock. This clock rate is a multiple of the natural |
| 201 |
H.261 frame rate (i.e. 30000/1001 or approx. 29.97 Hz). |
| 202 |
That way, for each frame time, the clock is just |
| 203 |
incremented by the multiple and this removes inaccuracy |
| 204 |
in calculating the timestamp. Furthermore, the initial |
| 205 |
value of the timestamp is random (unpredictable) to make |
| 206 |
known-plaintext attacks on encryption more difficult, see |
| 207 |
RTP [<a href="#ref-1" title=""RTP: A Transport Protocol for Real-Time Applications"">1</a>]. Note that if multiple frames are encoded in a |
| 208 |
packet (e.g. when there are very little changes between |
| 209 |
two images), it is necessary to calculate display times |
| 210 |
for the frames after the first using the timing |
| 211 |
information in the H.261 frame header. This is required |
| 212 |
because the RTP timestamp only gives the display time of |
| 213 |
the first frame in the packet. |
| 214 |
|
| 215 |
- The marker bit of the RTP header is set to one in the |
| 216 |
last packet of a video frame, and otherwise, must be |
| 217 |
|
| 218 |
|
| 219 |
|
| 220 |
<span class="grey">Turletti & Huitema Standards Track [Page 4]</span></pre> |
| 221 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span> |
| 222 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 223 |
|
| 224 |
|
| 225 |
zero. Thus, it is not necessary to wait for a following |
| 226 |
packet (which contains the start code that terminates the |
| 227 |
current frame) to detect that a new frame should be |
| 228 |
displayed. |
| 229 |
|
| 230 |
The H.261 data will follow the RTP header, as in: |
| 231 |
|
| 232 |
0 1 2 3 |
| 233 |
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 |
| 234 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 235 |
. . |
| 236 |
. RTP header . |
| 237 |
. . |
| 238 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 239 |
| H.261 header | |
| 240 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 241 |
| H.261 stream ... . |
| 242 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 243 |
|
| 244 |
The H.261 header is defined as following: |
| 245 |
|
| 246 |
0 1 2 3 |
| 247 |
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 |
| 248 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 249 |
|SBIT |EBIT |I|V| GOBN | MBAP | QUANT | HMVD | VMVD | |
| 250 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 251 |
|
| 252 |
The fields in the H.261 header have the following meanings: |
| 253 |
|
| 254 |
Start bit position (SBIT): 3 bits |
| 255 |
Number of most significant bits that should be ignored |
| 256 |
in the first data octet. |
| 257 |
|
| 258 |
End bit position (EBIT): 3 bits |
| 259 |
Number of least significant bits that should be ignored |
| 260 |
in the last data octet. |
| 261 |
|
| 262 |
INTRA-frame encoded data (I): 1 bit |
| 263 |
Set to 1 if this stream contains only INTRA-frame coded |
| 264 |
blocks. Set to 0 if this stream may or may not contain |
| 265 |
INTRA-frame coded blocks. The sense of this bit may not |
| 266 |
change during the course of the RTP session. |
| 267 |
|
| 268 |
Motion Vector flag (V): 1 bit |
| 269 |
Set to 0 if motion vectors are not used in this stream. |
| 270 |
Set to 1 if motion vectors may or may not be used in |
| 271 |
this stream. The sense of this bit may not change during |
| 272 |
the course of the session. |
| 273 |
|
| 274 |
|
| 275 |
|
| 276 |
<span class="grey">Turletti & Huitema Standards Track [Page 5]</span></pre> |
| 277 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span> |
| 278 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 279 |
|
| 280 |
|
| 281 |
GOB number (GOBN): 4 bits |
| 282 |
Encodes the GOB number in effect at the start of the |
| 283 |
packet. Set to 0 if the packet begins with a GOB header. |
| 284 |
|
| 285 |
Macroblock address predictor (MBAP): 5 bits |
| 286 |
Encodes the macroblock address predictor (i.e. the last |
| 287 |
MBA encoded in the previous packet). This predictor ranges |
| 288 |
from 0-32 (to predict the valid MBAs 1-33), but because |
| 289 |
the bit stream cannot be fragmented between a GOB header |
| 290 |
and MB 1, the predictor at the start of the packet can |
| 291 |
never be 0. Therefore, the range is 1-32, which is biased |
| 292 |
by -1 to fit in 5 bits. For example, if MBAP is 0, the |
| 293 |
value of the MBA predictor is 1. Set to 0 if the packet |
| 294 |
begins with a GOB header. |
| 295 |
|
| 296 |
Quantizer (QUANT): 5 bits |
| 297 |
Quantizer value (MQUANT or GQUANT) in effect prior to the |
| 298 |
start of this packet. Set to 0 if the packet begins with |
| 299 |
a GOB header. |
| 300 |
|
| 301 |
Horizontal motion vector data (HMVD): 5 bits |
| 302 |
Reference horizontal motion vector data (MVD). Set to 0 |
| 303 |
if V flag is 0 or if the packet begins with a GOB header, |
| 304 |
or when the MTYPE of the last MB encoded in the previous |
| 305 |
packet was not MC. HMVD is encoded as a 2's complement |
| 306 |
number, and `10000' corresponding to the value -16 is |
| 307 |
forbidden (motion vector fields range from +/-15). |
| 308 |
|
| 309 |
Vertical motion vector data (VMVD): 5 bits |
| 310 |
Reference vertical motion vector data (MVD). Set to 0 if |
| 311 |
V flag is 0 or if the packet begins with a GOB header, or |
| 312 |
when the MTYPE of the last MB encoded in the previous |
| 313 |
packet was not MC. VMVD is encoded as a 2's complement |
| 314 |
number, and `10000' corresponding to the value -16 is |
| 315 |
forbidden (motion vector fields range from +/-15). |
| 316 |
|
| 317 |
Note that the I and V flags are hint flags, i.e. they can be inferred |
| 318 |
from the bit stream. They are included to allow decoders to make |
| 319 |
optimizations that would not be possible if these hints were not |
| 320 |
provided before bit stream was decoded. Therefore, these bits cannot |
| 321 |
change for the duration of the stream. A conformant implementation |
| 322 |
can always set V=1 and I=0. |
| 323 |
|
| 324 |
<span class="h3"><a class="selflink" id="section-4.2" href="#section-4.2">4.2</a>. Recommendations for operation with hardware codecs</span> |
| 325 |
|
| 326 |
Packetizers for hardware codecs can trivially figure out GOB |
| 327 |
boundaries using the GOB-start pattern included in the H.261 data. |
| 328 |
(Note that software encoders already know the boundaries.) The |
| 329 |
|
| 330 |
|
| 331 |
|
| 332 |
<span class="grey">Turletti & Huitema Standards Track [Page 6]</span></pre> |
| 333 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span> |
| 334 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 335 |
|
| 336 |
|
| 337 |
cheapest packetization implementation is to packetize at the GOB |
| 338 |
level all the GOBs that fit in a packet. But when a GOB is too |
| 339 |
large, the packetizer has to parse it to do MB fragmentation. (Note |
| 340 |
that only the Huffman encoding must be parsed and that it is not |
| 341 |
necessary to fully decompress the stream, so this requires relatively |
| 342 |
little processing; example implementations can be found in some |
| 343 |
public H.261 codecs such as IVS [<a href="#ref-4" title=""rodeo/ivs/last_version"">4</a>] and VIC [<a href="#ref-9" title=""conferencing/vic"">9</a>].) It is recommended |
| 344 |
that MB level fragmentation be used when feasible in order to obtain |
| 345 |
more efficient packetization. Using this fragmentation scheme reduces |
| 346 |
the output packet rate and therefore reduces the overhead. |
| 347 |
|
| 348 |
At the receiver, the data stream can be depacketized and directed to |
| 349 |
a hardware codec's input. If the hardware decoder operates at a |
| 350 |
fixed bit rate, synchronization may be maintained by inserting the |
| 351 |
stuffing pattern between MBs (i.e., between packets) when the packet |
| 352 |
arrival rate is slower than the bit rate. |
| 353 |
|
| 354 |
<span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>. Packet loss issues</span> |
| 355 |
|
| 356 |
On the Internet, most packet losses are due to network congestion |
| 357 |
rather than transmission errors. Using UDP, no mechanism is available |
| 358 |
at the sender to know if a packet has been successfully received. It |
| 359 |
is up to the application, i.e. coder and decoder, to handle the |
| 360 |
packet loss. Each RTP packet includes a a sequence number field which |
| 361 |
can be used to detect packet loss. |
| 362 |
|
| 363 |
H.261 uses the temporal redundancy of video to perform compression. |
| 364 |
This differential coding (or INTER-frame coding) is sensitive to |
| 365 |
packet loss. After a packet loss, parts of the image may remain |
| 366 |
corrupt until all corresponding MBs have been encoded in INTRA-frame |
| 367 |
mode (i.e. encoded independently of past frames). There are several |
| 368 |
ways to mitigate packet loss: |
| 369 |
|
| 370 |
(1) One way is to use only INTRA-frame encoding and MB level |
| 371 |
conditional replenishment. That is, only MBs that change |
| 372 |
(beyond some threshold) are transmitted. |
| 373 |
|
| 374 |
(2) Another way is to adjust the INTRA-frame encoding |
| 375 |
refreshment rate according to the packet loss observed by |
| 376 |
the receivers. The H.261 recommendation specifies that a |
| 377 |
MB is INTRA-frame encoded at least every 132 times it is |
| 378 |
transmitted. However, the INTRA-frame refreshment rate |
| 379 |
can be raised in order to speed the recovery when the |
| 380 |
measured loss rate is significant. |
| 381 |
|
| 382 |
(3) The fastest way to repair a corrupted image is to request |
| 383 |
an INTRA-frame coded image refreshment after a packet |
| 384 |
loss is detected. One means to accomplish this is for the |
| 385 |
|
| 386 |
|
| 387 |
|
| 388 |
<span class="grey">Turletti & Huitema Standards Track [Page 7]</span></pre> |
| 389 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span> |
| 390 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 391 |
|
| 392 |
|
| 393 |
decoder to send to the coder a list of packets lost. The |
| 394 |
coder can decide to encode every MB of every GOB of the |
| 395 |
following video frame in INTRA-frame mode (i.e. Full |
| 396 |
INTRA-frame encoded), or if the coder can deduce from the |
| 397 |
packet sequence numbers which MBs were affected by the |
| 398 |
loss, it can save bandwidth by sending only those MBs in |
| 399 |
INTRA-frame mode. This mode is particularly efficient in |
| 400 |
point-to-point connection or when the number of decoders |
| 401 |
is low. The next section specifies how the refresh |
| 402 |
function may be implemented. |
| 403 |
|
| 404 |
Note that the method (1) is currently implemented in the VIC |
| 405 |
videoconferencing software [<a href="#ref-9" title=""conferencing/vic"">9</a>]. Methods (2) and (3) are currently |
| 406 |
implemented in the IVS videoconferencing software [<a href="#ref-4" title=""rodeo/ivs/last_version"">4</a>]. |
| 407 |
|
| 408 |
<span class="h3"><a class="selflink" id="section-5.1" href="#section-5.1">5.1</a>. Use of optional H.261-specific control packets</span> |
| 409 |
|
| 410 |
This specification defines two H.261-specific RTCP control packets, |
| 411 |
"Full INTRA-frame Request" and "Negative Acknowledgement", described |
| 412 |
in the next section. Their purpose is to speed up refreshment of the |
| 413 |
video in those situations where their use is feasible. Support of |
| 414 |
these H.261-specific control packets by the H.261 sender is optional; |
| 415 |
in particular, early experiments have shown that the usage of this |
| 416 |
feature could have very negative effects when the number of sites is |
| 417 |
very large. Thus, these control packets should be used with caution. |
| 418 |
|
| 419 |
The H.261-specific control packets differ from normal RTCP packets in |
| 420 |
that they are not transmitted to the normal RTCP destination |
| 421 |
transport address for the RTP session (which is often a multicast |
| 422 |
address). Instead, these control packets are sent directly via |
| 423 |
unicast from the decoder to the coder. The destination port for |
| 424 |
these control packets is the same port that the coder uses as a |
| 425 |
source port for transmitting RTP (data) packets. Therefore, these |
| 426 |
packets may be considered "reverse" control packets. |
| 427 |
|
| 428 |
As a consequence, these control packets may only be used when no RTP |
| 429 |
mixers or translators intervene in the path from the coder to the |
| 430 |
decoder. If such intermediate systems do intervene, the address of |
| 431 |
the coder would no longer be present as the network-level source |
| 432 |
address in packets received by the decoder, and in fact, it might not |
| 433 |
be possible for the decoder to send packets directly to the coder. |
| 434 |
|
| 435 |
Some reliable multicast protocols use similar NACK control packets |
| 436 |
transmitted over the normal multicast distribution channel, but they |
| 437 |
typically use random delays to prevent a NACK implosion problem [<a href="#ref-2" title="Don Towsley and James F. Kurose">2</a>]. |
| 438 |
The goal of such protocols is to provide reliable multicast packet |
| 439 |
delivery at the expense of delay, which is appropriate for |
| 440 |
applications such as a shared whiteboard. |
| 441 |
|
| 442 |
|
| 443 |
|
| 444 |
<span class="grey">Turletti & Huitema Standards Track [Page 8]</span></pre> |
| 445 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span> |
| 446 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 447 |
|
| 448 |
|
| 449 |
On the other hand, interactive video transmission is more sensitive |
| 450 |
to delay and does not require full reliability. For video |
| 451 |
applications it is more effective to send the NACK control packets as |
| 452 |
soon as possible, i.e. as soon as a loss is detected, without adding |
| 453 |
any random delays. In this case, multicasting the NACK control |
| 454 |
packets would generate useless traffic between receivers since only |
| 455 |
the coder will use them. But this method is only effective when the |
| 456 |
number of receivers is small. e.g. in IVS [<a href="#ref-4" title=""rodeo/ivs/last_version"">4</a>] the H.261 specific |
| 457 |
control packets are used only in point-to-point connections or in |
| 458 |
point-to-multipoint connections when there are less than 10 |
| 459 |
participants in the conference. |
| 460 |
|
| 461 |
<span class="h3"><a class="selflink" id="section-5.2" href="#section-5.2">5.2</a>. H.261 control packets definition</span> |
| 462 |
|
| 463 |
<span class="h4"><a class="selflink" id="section-5.2.1" href="#section-5.2.1">5.2.1</a>. Full INTRA-frame Request (FIR) packet</span> |
| 464 |
|
| 465 |
0 1 2 3 |
| 466 |
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 |
| 467 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 468 |
|V=2|P| MBZ | PT=RTCP_FIR | length | |
| 469 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 470 |
| SSRC | |
| 471 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 472 |
|
| 473 |
This packet indicates that a receiver requires a full encoded image |
| 474 |
in order to either start decoding with an entire image or to refresh |
| 475 |
its image and speed the recovery after a burst of lost packets. The |
| 476 |
receiver requests the source to force the next image in full "INTRA- |
| 477 |
frame" coding mode, i.e. without using differential coding. The |
| 478 |
various fields are defined in the RTP specification [<a href="#ref-1" title=""RTP: A Transport Protocol for Real-Time Applications"">1</a>]. SSRC is the |
| 479 |
synchronization source identifier for the sender of this packet. The |
| 480 |
value of the packet type (PT) identifier is the constant RTCP_FIR |
| 481 |
(192). |
| 482 |
|
| 483 |
<span class="h4"><a class="selflink" id="section-5.2.2" href="#section-5.2.2">5.2.2</a>. Negative ACKnowledgements (NACK) packet</span> |
| 484 |
|
| 485 |
The format of the NACK packet is as follow: |
| 486 |
|
| 487 |
0 1 2 3 |
| 488 |
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 |
| 489 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 490 |
|V=2|P| MBZ | PT=RTCP_NACK | length | |
| 491 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 492 |
| SSRC | |
| 493 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 494 |
| FSN | BLP | |
| 495 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 496 |
|
| 497 |
|
| 498 |
|
| 499 |
|
| 500 |
<span class="grey">Turletti & Huitema Standards Track [Page 9]</span></pre> |
| 501 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span> |
| 502 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 503 |
|
| 504 |
|
| 505 |
The various fields T, P, PT, length and SSRC are defined in the RTP |
| 506 |
specification [<a href="#ref-1" title=""RTP: A Transport Protocol for Real-Time Applications"">1</a>]. The value of the packet type (PT) identifier is |
| 507 |
the constant RTCP_NACK (193). SSRC is the synchronization source |
| 508 |
identifier for the sender of this packet. |
| 509 |
|
| 510 |
The two remaining fields have the following meanings: |
| 511 |
|
| 512 |
First Sequence Number (FSN): 16 bits |
| 513 |
Identifies the first sequence number lost. |
| 514 |
|
| 515 |
Bitmask of following lost packets (BLP): 16 bits |
| 516 |
A bit is set to 1 if the corresponding packet has been lost, |
| 517 |
and set to 0 otherwise. BLP is set to 0 only if no packet |
| 518 |
other than that being NACKed (using the FSN field) has been |
| 519 |
lost. BLP is set to 0x00001 if the packet corresponding to |
| 520 |
the FSN and the following packet have been lost, etc. |
| 521 |
|
| 522 |
<span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>. Security Considerations</span> |
| 523 |
|
| 524 |
Security issues are not discussed in this memo. |
| 525 |
|
| 526 |
Authors' Addresses |
| 527 |
|
| 528 |
Thierry Turletti |
| 529 |
INRIA - RODEO Project |
| 530 |
2004 route des Lucioles |
| 531 |
BP 93, 06902 Sophia Antipolis |
| 532 |
FRANCE |
| 533 |
|
| 534 |
EMail: [email protected] |
| 535 |
|
| 536 |
|
| 537 |
Christian Huitema |
| 538 |
MCC 1J236B Bellcore |
| 539 |
445 South Street |
| 540 |
Morristown, NJ 07960-6438 |
| 541 |
|
| 542 |
EMail: [email protected] |
| 543 |
|
| 544 |
Acknowledgements |
| 545 |
|
| 546 |
This memo is based on discussion within the AVT working group chaired |
| 547 |
by Stephen Casner. Steve McCanne, Stephen Casner, Ronan Flood, Mark |
| 548 |
Handley, Van Jacobson, Henning G. Schulzrinne and John Wroclawski |
| 549 |
provided valuable comments. Stephen Casner and Steve McCanne also |
| 550 |
helped greatly with getting this document into readable form. |
| 551 |
|
| 552 |
|
| 553 |
|
| 554 |
|
| 555 |
|
| 556 |
<span class="grey">Turletti & Huitema Standards Track [Page 10]</span></pre> |
| 557 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span> |
| 558 |
<span class="grey"><a href="./rfc2032">RFC 2032</a> RTP Payload Format for H.261 Video October 1996</span> |
| 559 |
|
| 560 |
|
| 561 |
References |
| 562 |
|
| 563 |
[<a id="ref-1">1</a>] Schulzrinne, H., Casner, S., Frederick, R., and |
| 564 |
V. Jacobson, "RTP: A Transport Protocol for Real-Time |
| 565 |
Applications", <a href="./rfc1889">RFC 1889</a>, January 1996. |
| 566 |
|
| 567 |
[<a id="ref-2">2</a>] Sridhar Pingali, Don Towsley and James F. Kurose, A |
| 568 |
comparison of sender-initiated and receiver-initiated |
| 569 |
reliable multicast protocols, IEEE GLOBECOM '94. |
| 570 |
|
| 571 |
[<a id="ref-3">3</a>] Thierry Turletti, H.261 software codec for |
| 572 |
videoconferencing over the Internet INRIA Research Report |
| 573 |
no 1834, January 1993. |
| 574 |
|
| 575 |
[<a id="ref-4">4</a>] Thierry Turletti, INRIA Videoconferencing tool (IVS), |
| 576 |
available by anonymous ftp from zenon.inria.fr in the |
| 577 |
"rodeo/ivs/last_version" directory. See also URL |
| 578 |
<<a href="http://www.inria.fr/rodeo/ivs.html">http://www.inria.fr/rodeo/ivs.html</a>>. |
| 579 |
|
| 580 |
[<a id="ref-5">5</a>] Frame structure for Audiovisual Services for a 64 to 1920 |
| 581 |
kbps Channel in Audiovisual Services ITU-T (International |
| 582 |
Telecommunication Union - Telecommunication |
| 583 |
Standardisation Sector) Recommendation H.221, 1990. |
| 584 |
|
| 585 |
[<a id="ref-6">6</a>] Video codec for audiovisual services at p x 64 kbit/s |
| 586 |
ITU-T (International Telecommunication Union - |
| 587 |
Telecommunication Standardisation Sector) Recommendation |
| 588 |
H.261, 1993. |
| 589 |
|
| 590 |
[<a id="ref-7">7</a>] Digital Methods of Transmitting Television Information |
| 591 |
ITU-R (International Telecommunication Union - |
| 592 |
Radiocommunication Standardisation Sector) Recommendation |
| 593 |
601, 1986. |
| 594 |
|
| 595 |
[<a id="ref-8">8</a>] M.A Sasse, U. Bilting, C-D Schulz, T. Turletti, Remote |
| 596 |
Seminars through MultiMedia Conferencing: Experiences |
| 597 |
from the MICE project, Proc. INET'94/JENC5, Prague, June |
| 598 |
1994, pp. 251/1-251/8. |
| 599 |
|
| 600 |
[<a id="ref-9">9</a>] Steve MacCanne, Van Jacobson, VIC Videoconferencing tool, |
| 601 |
available by anonymous ftp from ee.lbl.gov in the |
| 602 |
"conferencing/vic" directory. |
| 603 |
|
| 604 |
|
| 605 |
|
| 606 |
|
| 607 |
|
| 608 |
|
| 609 |
|
| 610 |
|
| 611 |
|
| 612 |
Turletti & Huitema Standards Track [Page 11] |
| 613 |
</pre> |